Google LLC v Defteros

KIEFEL CJ AND GLEESON J. The respondent, George Defteros, is a solicitor who has practised criminal law for many years. In the course of that practice he has acted for persons who became well-known during Melbourne's "gangland wars", including Dominic ("Mick") Gatto and Mario Condello. In 2004, the respondent and Mr Condello were charged with conspiracy to murder and incitement to murder Carl Williams and others and were committed to stand trial. In 2005, the Director of Public Prosecutions withdrew the charges against the respondent. In the intervening period the prosecution of the respondent and Mr Condello was widely reported, including in The Age newspaper, and articles were placed on that newspaper's website.

The appellant, Google LLC, makes available to those seeking to navigate information on the World Wide Web its search engine. In early 2016, the respondent became aware that an internet search of his name using the Google search engine produced search results which included a snippet of an article published by The Age in 2004, on the day after the respondent was charged. This was termed the "Search Result" in the proceedings below. The title of the article, displayed in the Search Result, contained a hyperlink to the full article on The Age's website. The article was entitled "Underworld loses valued friend at court". In the proceedings below it was referred to as the "Underworld article". Together, the Search Result and the Underworld article were said to comprise the "Web Matter" which the respondent claimed defamed him.

In proceedings brought in the Supreme Court of Victoria, the respondent claimed damages for defamation from the appellant as publisher of the Web Matter. The appellant denied publication. It relevantly pleaded, in the alternative, both the common law and statutory defences of innocent dissemination and qualified privilege.

The trial judge, Richards J, found that the appellant had published the Web Matter, based on her Honour's view of the significance of the insertion of a hyperlink to The Age website in the Search Result. Her Honour found that the Web Matter conveyed the defamatory imputation that the respondent had crossed the line from being a professional lawyer for, to become a confidant and friend of, criminal elements. Only the defence of statutory qualified privilege was made out and only with respect to a substantial proportion, but not all, of the persons to whom her Honour found the Web Matter had been published. Her Honour awarded the respondent general damages of $40,000 and later made further orders for interest and costs, the latter reflecting the "mixed success" of the parties in the proceedings.

The action concerning the Web Matter was heard by the trial judge together with an action brought by the respondent with respect to other material which it was alleged was later published by the appellant and which was also said to be defamatory of him. That second action was dismissed.

Relevantly to this appeal, the appellant sought leave from the Court of Appeal to appeal the judgment in respect of the Web Matter; and the respondent sought leave to cross-appeal in relation to the costs order for the same. The parties also filed applications and cross-applications for leave to appeal in respect of the second action. The Court of Appeal (Beach, Kaye and Niall JJA) granted both applications for leave to appeal but dismissed the appeals. It also dismissed both cross-applications for leave to appeal in relation to costs. The appeal to this Court

The appellant appeals to this Court from that part of the judgment given by the Court of Appeal on 17 June 2021 concerning the Web Matter, pursuant to a grant of special leave to appeal which was conditional upon the appellant paying the respondent's costs of the appeal and not seeking to disturb the costs orders in the courts below.

The appellant's original Notice of Appeal contains three grounds. The first and principal ground is that the Court of Appeal was wrong to conclude that the appellant published the Web Matter. The second and third are that the Court of Appeal was wrong to reject the appellant's defences of common law and statutory qualified privilege. At the outset of the hearing of this appeal, the appellant was given leave to amend its Notice of Appeal to add a further ground – that the Court of Appeal was wrong to have rejected its defence of innocent dissemination at common law and pursuant to s 32 of the Defamation Act 2005 (Vic).

It will not be necessary to consider the grounds relating to the defences. Applying the settled principles of the common law of defamation concerning the communication of defamatory matter and participation in it, it cannot be concluded that the appellant, by providing the hyperlink, published the Web Matter. The operation of the appellant's search engine

In essence, an enquiry made using the appellant's search engine may elicit a number of results in response which, for each entry, have in common these features: the title of the webpage referred to in the entry; the Uniform Resource Locator ("URL") associated with the webpage; and a snippet of the content from the webpage (or an image). A hyperlink is provided in the title. When the user clicks on it, the webpage appears on the user's web browser.

The trial judge described the operation of, and results obtained from, the appellant's search engine in the context of the World Wide Web by reference to the evidence given at the trial. The accuracy of her Honour's description was not disputed by the parties and the description was adopted by the Court of Appeal. It is convenient to take the same course. Her Honour explained that: "The World Wide Web is a vast system of linked documents accessed by the internet. It comprises trillions of pages generated by millions of people and organisations worldwide, and is constantly changing and expanding. Each webpage has a unique Uniform Resource Locator or URL, which acts as an address for that webpage. The code commonly used for creating webpages is called HyperText Markup Language or HTML. The HTML code of a webpage is read by a software program called a browser, which displays text and images to a user on the user's device. A webpage typically contains hyperlinks to other webpages. A hyperlink is some HTML code that contains a URL for another webpage. When a user clicks on a hyperlink on one webpage, the browser on the user's computer displays text and images from the other webpage. It is the myriad of hyperlinks between webpages that makes the Web what it is today. The Web is something like a constantly growing, rapidly changing, vast digital library, and navigating it can be a challenge. It is easy enough to find a particular website if one knows the URL for the website. A user can type that URL into the browser's address bar, and can then follow hyperlinks within the site or use the site's search function. Similarly, a user who knows the exact URL for a specific webpage can type that in to the browser's address bar, and be taken directly to the webpage. However, users who do not know the URL for the webpage containing the information they seek need other tools to find their way around. One popular tool for locating information on the Web is a search engine, which is an automated information retrieval system designed to allow a user to navigate the extensive information on the Web by the use of user-designed queries. The Google search engine is one of a number of search engines that are available; two other well-known search engines are Yahoo! and Bing. A user of the Google search engine can use it to search the entire Web, using Google Web Search. Google makes other more specific search functions available, for searching images, maps, videos, news, flights, and books. Search engines use computer algorithms to make predictions about what webpages among the trillions of pages constituting the Web are most likely to be of interest to a user, by responding to the user's search query. Typically, a search engine returns a list of results in response to a search query, in the form of hyperlinks to webpages on the Web. The Google search engine does this by first identifying what information is available on the Web, using a Web crawler program. This fully automated program uses a large number of computers to constantly visit and process webpages on the Web. The crawler program determines which websites to crawl and how often, as well as what information is collected from them. Webpages that it ranks as important are crawled for new data more frequently than less important pages. Every time a webpage is re‑crawled and new data is detected, the stored data relating to that webpage is updated. Next, the Google search engine uses an indexing program to organise the data obtained by the Web crawler into a form that is more easily searched by computer algorithms. The indexer program builds a list of every webpage that contains each word found during the crawling stage. The indexer program also notes other aspects of a webpage, such as the date it was published, whether the page comprises text, images or video, and whether the webpage appears to be a news article. The resulting index contains each word and a list of the unique IDs that relate to the webpages that contain that word. The index is constantly refreshed as the Web crawler detects new data on the Web. Then, when a user enters a search query in Google Web Search, the words from that query are evaluated by a series of algorithms, against the information in the index, as it is at that precise point in time. The search result that Google presents to the user is a list of links to webpages, ranked according to relevance, as estimated by the ranking algorithm. The ranking program uses various 'signals' or clues to identify what results the user is most likely looking for. Some of the signals used by the Google search engine are: (a) the number of times one or more of the user's search terms appear on the webpage, as indexed by the indexer program; (b) how often other web pages link to that web page, and the importance of the linking webpages (this signal is known as PageRank); (c) how recently the content of that webpage was published or updated (freshness); (d) evaluating the order in which the search terms appear on a webpage; (e) the location of the user, as determined from the user's Internet Protocol or IP address; and (f) the user's previous search history. The format of the search results that Google presents to a user may vary depending on the user's device and browser. Typically, for each webpage listed in the search result, the user is shown the title of the webpage, with the search terms in bold. The title is also a hyperlink, which the user can click to reach the webpage. The result also contains a 'snippet' of the content of the webpage, and a shortened form of its URL. There may be many pages of search results, with the webpages that the algorithm determines to be most relevant appearing on the first page of results. A Google image search results in a display of a collection of image 'thumbnails', each of which contains a hyperlink to an interstitial page that provides more information about the image. A click on the interstitial page then takes the user to the webpage where the image was found by the Google search engine. The Google search engine conducts a Web search almost instantaneously – the average search response time is about half a second. Every month, over 100 billion searches are made by users of the Google search engine. Of these, 15% or more than 500 million searches each day are searches that have not been made using the Google search engine before. Although humans who work for Google design the search engine and its component programs, a Google search is a fully automated process that operates without human intervention. The Google search engine is not capable of evaluating the meanings conveyed by the words and images on a webpage, including whether they are true, false, or defamatory."

Her Honour went on to refer to the appellant's ability to remove a webpage, identified by its URL, from the search results. It is not necessary to discuss this or the appellant's policy relating to removal. They are not relevant to the question whether a matter which is defamatory is published by what is provided by the Search Result, namely the provision of a hyperlink with accompanying text that is not itself defamatory.

The use of the appellant's search engine, using the respondent's name as the search query, yielded a set of search results which were displayed on the appellant's webpage. The results included the Search Result the subject of the Web Matter: "Underworld loses valued friend at court ‑SpecialsGanglandKillings ... www.theage.com.au > Features > Crime & Corruption ▼ June 18 2004 ‑ Pub bouncer‑turned‑criminal lawyer George Defteros always prided himself on being able to avoid a king hit – The Age Online"

If the person conducting the search clicked the hyperlink in the title in the Search Result, the Underworld article and a photograph of the respondent would be displayed. Publication – reasons of the courts below

The trial judge concluded that the appellant "publishes a webpage that is reached by clicking on a hyperlink within a search result, because its provision of a hyperlinked search result is instrumental to the communication of the content of the webpage to the user. The Google search engine lends assistance to the publication of the content of a webpage on the user's device, by enabling the user to enter a search query and, a few clicks later, to view content that is relevant to the user's search." That is what the appellant's search engine is designed to do, her Honour said. This conclusion was said to follow from an application of the principles in Webb v Bloch. It followed that the appellant published the Web Matter by providing a hyperlink in the Search Result to the Underworld article.

Her Honour further held that, as a secondary publisher, the appellant was not liable for publication of the defamatory matter complained of until a reasonable time after it had been notified that the Search Result included that matter.

It would appear that in argument before her Honour, an analogy was sought to be drawn between search results and a catalogue to a library collection. Her Honour's findings referred to the World Wide Web as a "vast digital library". Her Honour considered that whilst the analogy was useful, it did not quite capture what occurred. In her Honour's view, a hyperlink is more than simply a reference to where information can be found on the Web. A closer analogy is a librarian who fetches a book and delivers it to the user bookmarked at the relevant page. All that is left for the user to do is to open the book and read it. Her Honour said: "[i]n my view, the provision of a hyperlink within a search result facilitates the communication of the contents of the linked webpage to such a substantial degree that it amounts to publication of the webpage".

The Court of Appeal held there to be no error in the trial judge's conclusion that the appellant became a publisher of the Underworld article seven days after it received notification of the defamatory matters contained therein. Their Honours considered that the Search Result was an "enticement" to the reader to click on the hyperlink to obtain more information about the respondent. In doing so, their Honours expressly approved the approach of Hinton J in Google Inc v Duffy. Their Honours also agreed with the approach of Kourakis CJ in Google Inc v Duffy and considered that the Search Result, containing the hyperlink to the Underworld article, "incorporated" the content of the Underworld article even though it did not repeat any of the text in the article said to be defamatory of the respondent. Both the concepts of "enticement" and "incorporation" were said to reflect the test in Webb v Bloch because they fasten on steps that lend assistance to the publication. The combination of the search terms, the text of the Search Result and the insertion of the hyperlink both directed and encouraged the reader to click on the link for further information. The publication of defamatory matter – principles

In Trkulja v Google LLC, it was said that "[i]n point of principle, the law as to publication is tolerably clear". This statement was made by reference to Webb v Bloch. In Fairfax Media Publications Pty Ltd v Voller, a decision which was given subsequent to the judgment appealed from in this case, a majority of this Court affirmed the principles relating to the publication of defamatory matter which had been stated in Webb v Bloch.

The issue in Voller was what was meant by the requirement of the common law of defamation that the publication of defamatory matter must be intentional. It was held that all that is required is that the defendant's act of participation in publication be voluntary.

The majority in Voller also considered more broadly what the law requires for there to be a publication and for a person to be liable as a publisher. Publication was explained as the actionable wrong in the tort of defamation by which harm is occasioned to a person's reputation. Publication was described as a technical term, which is to be understood as a bilateral act by which the publisher makes the defamatory material available and a third party has it available for their comprehension. It may be understood as the process by which a defamatory statement or imputation is conveyed. Adopting Webb v Bloch, any act of participation in the communication of defamatory matter to a third party is sufficient to make a defendant a publisher, regardless of their knowledge or intent. So understood, a person who has been instrumental in, or contributes to any extent to, the publication of defamatory matter is a publisher. Applying principles to facts

The difficulty in a case such as the present does not arise from any uncertainty as to the principles concerning the publication of defamatory matter. As the Court went on to observe in Trkulja, the difficulty which can arise is in the application of the principles to the particular facts of the case. Difficulty may arise where the principles relating to publication are considered in the abstract and not in the factual context in which they are stated.

It was not suggested by the courts below that the appellant, as an internet search engine operator, actually communicated the defamatory material. It is of course possible that search results may themselves contain matter which is defamatory. This was acknowledged in Trkulja. But that is not this case.

The question which arises here is whether providing search results which, in response to an enquiry, direct the attention of a person to the webpage of another and assist them in accessing it amounts to an act of participation in the communication of defamatory matter. The broad rule applied

It must be accepted that the principles regarding publication of defamatory material have been regarded as having a broad reach. In Voller, reference was made to what had been said by Abella J in Crookes v Newton in this regard, namely that "the breadth of activity captured by the traditional publication rule is vast". But that is not to say that there are no limits to what actions are captured by the rule or that any connection between a person's act and the publication, however remote, will render them liable as a tortfeasor.

The correctness of the statement in Crookes v Newton may be seen most clearly from the cases which marked the early development of the defence of innocent dissemination, which was the common law's response to the strictness with which the rule of publication had been applied. But for the defence, persons who sold newspapers or periodicals would be liable as publishers, as would those who lent or sold copies of a book containing defamatory matter and persons who delivered parcels which contained libellous documents. What they may be seen to have in common is that the acts in question directly conveyed the defamatory material.

The acts in question in Webb v Bloch and Voller are different from the examples given above. They did not involve the actual distribution of the defamatory material. The defendants were involved in and participated in the publication of defamatory material by conduct which preceded its dissemination. In Webb v Bloch, the defendants approved the creation of the defamatory material for the purpose of its distribution and, in Voller, the defendants encouraged the creation of the alleged defamatory matter. In Voller the defendants additionally facilitated its publication by providing a platform for its communication.

In Webb v Bloch, the plaintiff and the defendants were members of committees which represented wheat growers and purchasers of wheat scrip in different States. The committees had as their common object to seek compensation from the Government of South Australia for the benefit of the growers. To that end, the committees each instituted and controlled various legal proceedings. A difference of view arose between the South Australian committee, which the plaintiff chaired, and the Victorian committee, of which most of the defendants were members, as to whether litigation controlled by the South Australian committee should be settled. A solicitor acted for both committees.

One of the defendants, Mr Bloch, instructed the solicitor to compose a circular to be published to the growers in South Australia with a view to soliciting funds for further litigation, an action which was approved by the other defendants. The circular contained information which was untrue and defamatory of the plaintiff, Mr Webb. Mr Bloch knew certain of the contents of the circular to be untrue, one defendant was not aware of the circular's contents but was aware that certain statements in it were untrue and the other defendants were unaware of either the contents of the circular or whether statements in it were true or false. The majority, Knox CJ and Isaacs J, held all defendants to be liable for the publication of the defamatory matter in the circular. The solicitor's malice was also attributed to them.

Isaacs J explained that "publication" in the law of libel did not just mean to physically distribute it. He said that "[t]o publish a libel is to convey by some means to the mind of another the defamatory sense embodied in [the document conveying the defamation]". On that basis the solicitor had "published" the defamation even though he had not himself disseminated it.

Isaacs J then went on to discuss the meaning of "publication" as relevant to the defendants. His Honour gave examples, by reference to a text and cases, of persons who were not the authors or disseminators of defamatory matter but were treated as publishers, as where a person suggested that the matter be written; caused it to be published; approved, concurred or showed their assent or gave their approbation to the libel; or assisted or encouraged the damage to another's reputation. Many of the cases drew upon the analogy of a principal and agent to explain the person's liability as publisher.

Isaacs J concluded that whilst it may be said that the solicitor was the "real author" or the "master mind", and the defendants were the "intermediate agents" to disseminate the libel, "[t]hey cannot employ the master mind for the very purpose, accept its suggestions, approve and disseminate its production, and then disclaim its malice".

The defendants in Voller were media companies which each maintained a public Facebook page on which they posted hyperlinks to news stories, with an associated headline, comment and image. Clicking on the hyperlink took the reader to the news story on the defendant's website. But it was not these acts which were said to involve the defendants in publication of the alleged defamatory material; rather it was what the defendants did in seeking commentary upon the articles which brought them within the principles stated in Webb v Bloch. The defendants were found to have invited and encouraged comment about the articles from Facebook users. It was the response by some third-party users to that encouragement which contained the alleged defamatory material. It was the defendants' acts in facilitating, encouraging and assisting the posting of comments by the third-party users which rendered them liable as publishers of those comments.

In each of Webb v Bloch and Voller the defendants' acts of approval and encouragement were clearly connected to the creation of the matter in question for the purpose of it being communicated to others. Voller had the additional feature that the media companies provided the place for publication. The present case has none of these features. The provision of the Search Result, including the hyperlink, has no connection to the creation of the Underworld article; its creation was in no way approved or encouraged by the appellant; and the appellant did not participate in it being placed on The Age's website. Closer analogies – US and Canadian cases

The circumstances of the present case are closer in kind to those considered in a line of decisions from the courts of the United States and Canada.

Klein v Biben concerned an article in a periodical which the plaintiff alleged was libellous. A later issue of the same periodical was alleged to have repeated the publication by the statement "For more details about [the plaintiff], see the Washington News Letter in The American Hebrew, May 12, 1944". That is to say, it drew attention to the earlier article. The question whether there was republication by reason of the specific reference in the later issue of the periodical was answered in the negative by the Court of Appeals of New York.

In MacFadden v Anthony a magazine available for purchase contained an article which the plaintiff claimed was libellous. A radio commentator called attention to the article in the course of a broadcast. It was not claimed that the commentator repeated the words of the publication, verbatim or in substance, or that any of the words spoken by him were themselves defamatory. On a motion to dismiss the complaint for legal insufficiency, the New York Supreme Court applied Klein and held that the commentator's statement was not a publication or republication of the libel.

Closer in time and fact is Carter v BC Federation of Foster Parents Association, where mention was made in a printed newsletter of the internet address of an internet forum which contained defamatory comments. The Court of Appeal of British Columbia, adopting Klein and MacFadden, held that a reference to an article which does not repeat the defamatory comment itself is not a republication of it.

These cases were referred to with approval by Abella J in Crookes v Newton, a decision of the Supreme Court of Canada. Mr Newton operated a website which contained commentary on issues such as free speech. One of the articles he posted on the website contained hyperlinks to other websites which contained information about Mr Crookes that were said to include matters defamatory of him. In his action brought against Mr Newton, Mr Crookes alleged that by providing the hyperlinks Mr Newton was publishing the defamatory matter.

The starting point for the analysis undertaken by Abella J (with whom Binnie, LeBel, Charron, Rothstein and Cromwell JJ concurred and McLachlin CJ and Fish J substantially agreed) was the statement of the rule that a defendant who, by any act, conveys defamatory meaning is a publisher and an acknowledgment of the breadth of actions which fall within this traditional publication rule. This accords with Australian case law, as does the observation by Deschamps J that publication has a bilateral nature.

It needs to be said at the outset that there are aspects of the reasons of Abella J which, in our view, either cannot or should not be followed by this Court. In the former category are factors such as the public interest in protecting freedom of expression and other values drawn from the Canadian Charter of Rights and Freedoms. In the latter are policy considerations such as the benefits to be gained from the dissemination of information over the Internet. It is preferable to apply settled principles regarding publication to a situation presented by new technology. The result is that internet search engine results that are not themselves defamatory do not come within the purview of publication and it is not necessary to consider whether those principles should be adapted. In any event, it is clear that Abella J made reference to these other factors and considerations as confirming the correctness of her Honour's essential reasoning. That reasoning is compelling.

In her Honour's view, hyperlinks "are, in essence, references". A hyperlink provides a reference to another source and does not itself constitute publication of it. Although the link may facilitate the transfer of information, which her Honour accepted was a hallmark of publication, it is equally clear that when a person follows a link, they are leaving one source and moving to another. "The ease with which the referenced content can be accessed does not change the fact that, by hyperlinking, an individual is referring the reader to other content."

Hyperlinks thus have the same relationship to the content to which they link as do references, her Honour observed. Both communicate that something exists but do not by themselves communicate its content. Both require an act on the part of a third party before access is gained to the content. The fact that accessing the content is made far easier with hyperlinks does not alter the fact that a hyperlink, by itself (and as distinct from a search result in which the link is embedded), is "content‑neutral".

McLachlin CJ and Fish J agreed in large part with the reasons of Abella J, but held that a hyperlink will constitute publication if, read contextually, the text that includes the hyperlink constitutes adoption or endorsement of the specific content to which it links. A mere reference without endorsement or adoption remains content‑neutral; but where the specific content is endorsed or adopted the reference can be understood to actually incorporate the defamatory content.

It will be recalled that the Court of Appeal in this case adopted the possibility of defamatory matter being published by way of incorporation by reference into a search result. In our view, this suffers from two difficulties. The first is that whereas incorporation by reference clearly has a place in contract law and other areas of law, such as patent law and the law relating to wills, it can have no place in the law of defamation, which requires that the defamatory meaning be conveyed for publication to be complete. The second is that in any event it is unnecessary. Cases such as Webb v Bloch show that the rules of publication apply where a person endorses, adopts or otherwise approves of defamatory matter which is to be published.

Deschamps J favoured a more nuanced approach, which requires consideration of factors such as whether the hyperlinks are "deep" or "shallow" and how they operate. This approach was not pursued by the respondent in this case and such factors were not the subject of findings by the trial judge. Publication?

It is not suggested that the appellant itself communicated the defamatory matter in the Underworld article, which appeared on The Age's website. Unlike the defendants in the innocent dissemination cases, the appellant did not do so by selling, distributing or otherwise disseminating the matter complained of. More relevant to this appeal, and by way of contrast with the circumstances of the defendants in those cases, is the hypothetical example of a person from whom directions are sought as to where a periodical might be obtained from a retail outlet. If that person gives directions or even escorts the enquirer to the place of sale, it could hardly be suggested that the person has communicated defamatory matter contained in the periodical. Nor could those actions be said to involve participation in the communication of such defamatory matter.

The publication of the defamatory matter here is best understood as occurring by the communication of the article on The Age's webpage to persons accessing the webpage. The questions raised by Webb v Bloch, Trkulja and Voller are whether the appellant could be said to have participated in the communication of the matter contained in the Underworld article or to have been instrumental in, or have contributed to any extent to, the publication of that matter. The phrases in question imply some connection on the part of a defendant's conduct with the acts which constitute publication of the defamatory matter. They do not suggest something which occurs that is physically remote from the acts constituting publication of the defamatory matter, and thus distinct from it. The observation of Abella J in Crookes v Newton, that by using a hyperlink a person moves from one webpage to another, serves to highlight that remoteness here.

It cannot be said that the appellant was involved in the communication of the defamatory material by reference to the circumstances in Webb v Bloch and Voller. It did not approve the writing of defamatory matter for the purpose of publication. It did not contribute to any extent to the publication of the Underworld article on The Age's webpage. It did not provide a forum or place where it could be communicated, nor did it encourage the writing of comment in response to the article which was likely to contain defamatory matter. Contrary to the finding of the trial judge, the appellant was not instrumental in communicating the Underworld article. It assisted persons searching the Web to find certain information and to access it.

As the trial judge found, navigating the Web can be a challenge. Search engines assist in that process. But the analogy drawn by her Honour between a search result and a librarian handing over to a library user a book marked at a particular page is problematic, not the least because a search result is only one of a number of responses to an enquiry, as was the case here. Properly understood, a search result conveys to the person searching that they may be interested in one or more of the results. The person is not directed to a particular result, as the Court of Appeal implied. According to the trial judge's findings, results are ranked by the use of an algorithm having regard to relevance using "signals" or clues as to what the person searching is looking for according to the words used in their enquiry. The search result merely refers, in the sense of drawing attention, to a webpage. As Abella J pointed out in Crookes v Newton, there is a difference between drawing a person's attention to the existence of an article and communicating its content. And whilst it may be said that the use of a hyperlink may mean The Age gains a reader, that does not make the appellant something other than a reference provider.

The respondent submitted that the Search Result had the added feature, to which the Court of Appeal referred, that by its terms it "enticed" the person searching to open the webpage. It is difficult to see how this level of excitement could be said to be generated by the words of the Search Result. Moreover, it needs to be borne in mind that the person has already activated a search for particular information before the result is received. As is the case with any search result, a person will employ the hyperlink if they think the webpage to which they are directed may provide the information they seek. It is notable that the respondent also submitted that the decisions of the New York courts should also be seen as involving "enticement". In doing so the respondent appears to equate "entice" with "call attention to".

As mentioned above, in Crookes v Newton Abella J pointed out that when a person chooses to click on the hyperlink, they leave the search result and go to a different webpage. In the course of argument on this appeal, another example was given of how a person might be referred or directed to another webpage that may contain defamatory material. A person might give to another the text and symbols that constitute the URL address for a webpage. It could hardly be suggested that, without more, the provision of a URL address is participation in the communication of defamatory matter which happens to be at that address. Yet the provision of a hyperlink involves little more. In reality, a hyperlink is merely a tool which enables a person to navigate to another webpage.

The question of whether the appellant could be said to participate comes down to the assistance provided by the hyperlink to move to another webpage. This is not a strong basis for liability and it finds no support in existing authority in Australia or recent cases elsewhere. As observed in Crookes v Newton, a hyperlink is content‑neutral. A search result is fundamentally a reference to something, somewhere else. Facilitating a person's access to the contents of another's webpage is not participating in the bilateral process of communicating its contents to that person. To hold that the provision of a hyperlink made the appellant a participant in the communication of the Underworld article would expand the principles relating to publication.

The aim or purpose of the appellant in making information universally accessible should not be confused with whether it is motivated to and does produce income by providing hyperlinks, conducting a business in that sense. Neither factor is relevant to whether there is a publication. That question focuses upon what the appellant in fact does.

For completeness, it should be noted that the respondent submitted that when notice was given to the appellant, as the internet search engine operator, that The Age's webpage contained defamatory material, the appellant's failure to remove the reference was relevant to the question of publication. The submission assumes that the appellant was a publisher, which, for the reasons given above, is not the case. It is also incorrect in principle to employ notification in connection with publication. Notice of the existence of defamatory matter may be relevant to knowledge in the defence of innocent dissemination, but it is not relevant to the question whether defamatory material is published. The submission bespeaks some notion of duty on the part of the appellant to the respondent which is breached. Orders

The following orders should be made: Appeal allowed. Set aside order 2 of the orders of the Court of Appeal of the Supreme Court of Victoria made on 17 June 2021 and, in lieu thereof, order that: (a) the appeal be allowed; (b) order 1 of the orders of the primary judge made on 6 May 2020 be set aside and, in lieu thereof, it be ordered that there is judgment for the defendant; and (c) order 1 of the orders of the primary judge made on 3 June 2020 be set aside. The appellant pay the respondent's costs of the appeal.

GAGELER J. Whether someone is a publisher of defamatory matter communicated by means of the Internet is determined in the application of the standard common law principles expounded by the majority in Fairfax Media Publications Pty Ltd v Voller by reference to Webb v Bloch (as reconfirmed in Trkulja v Google LLC) and Dow Jones & Co Inc v Gutnick.

Those standard common law principles posit that publication is a process which includes making matter available for comprehension by a third party (relevantly by including the matter on a webpage) and which is completed upon the third party having that matter available for comprehension (relevantly by viewing the webpage). Those principles further posit that a publisher is anyone who is an active and voluntary participant in any part of that process of publication.

Applying those principles, I agree with the conclusion reached by Kiefel CJ and Gleeson J that Google LLC was not a publisher of the defamatory matter contained in the article on a webpage on The Age website in the publication that occurred each time a third party who had entered "George Defteros" into the Google search engine clicked on the hyperlink in one of the search results so as to be connected to The Age website, where the article could be viewed. I also agree with the reasons given by Kiefel CJ and Gleeson J for reaching that conclusion.

Mindful that other members of this Court reach the opposite conclusion adhering to the same common law principles, I add the following observations.

The majority in Voller drew attention to the consistency of the common law principles there expounded with the common law principles expounded by the Supreme Court of Canada in Crookes v Newton. Despite calling attention to the fact that "the breadth of activity captured by the traditional publication rule is vast", Abella J (with whom five other Justices concurred) in Crookes applied those common law principles to conclude that a "mere" hyperlink constitutes no more than a "reference" to the existence and location of a webpage where matter can be viewed as distinct from a publication of the matter contained on that webpage. Abella J reasoned to that conclusion taking the view that a hyperlink by itself communicates that something exists, but does not communicate its content and, in that way, is "content-neutral".

That reasoning of Abella J in Crookes involved application to a hyperlink of the general understanding that to provide to a third party a reference to the location at which matter on a designated topic is available to be found does not at common law, without more, amount to publication of that matter to that third party. That understanding is not contradicted by anything in the reasoning of the majority in Voller and, as noted by Kiefel CJ and Gleeson J, is directly supported by the earlier New York case law analysed in Crookes. I cannot read the New York cases as having turned on the application of the "single publication rule" considered and rejected in Gutnick. Publication was emphasised in Voller to be a technical concept and was implicitly accepted in Voller to be used in no different sense in the common law of Australia from the sense in which it is used in the common law of New York. In that light, I read the New York cases analysed in Crookes as illustrations of a category of conduct – neutral referencing – that has been recognised not to amount to participation in a process of publication.

The conclusion in Crookes, that a mere hyperlink is to be characterised as neutral referencing, and therefore does not amount to participation in the process of publication of the matter to a third party who clicks on the hyperlink, is by no means incontestable. Of its nature, a hyperlink goes beyond merely indicating the existence and location of a webpage in that it also facilitates immediate access to that webpage. More than just indicating the location on the Internet where referenced matter is to be found, the hyperlink provides a shortcut to that location.

That being so, the circumstance that the conclusion in Crookes has been reached in the outworking of materially identical common law principles and has stood for more than a decade in Canada counts in favour of that conclusion being accepted and assimilated into the outworking of the applicable common law principles in Australia. To repeat a point made in Voller, with reference to Crookes, "given that a strength of common law reasoning lies in its ability to assimilate and build upon collective experience, when grappling with the application of common law principles of tortious liability inherited from a common source to meet common challenges presented by emerging global phenomena, 'convergence ... is preferable to divergence even if harmonisation is beyond reach'".

The ubiquity of the Internet and the centrality of hyperlinks to its operation make consistency in the characterisation of a hyperlink across common law jurisdictions especially desirable. As described by Matthew Collins, in language partially quoted by Abella J in Crookes: "Hyperlinks are the synapses connecting different parts of the world wide web. Without hyperlinks, the web would be like a library without a catalogue: full of information, but with no sure means of finding it. Almost every web page contains hyperlinked information, so that content is endlessly connected to other content."

To accept that the provision of a hyperlink is not enough to amount to participation in the process of publication which is completed when a third party clicks on the hyperlink so as to view the webpage, however, is not to deny that the provision of a hyperlink might combine with other factors to amount to participation in that process of publication of matter on that other webpage. The late 19th century decision of the English Court of Appeal in Hird v Wood – treating as a publisher a man who sat on a stool smoking a pipe and continuously pointing to a placard so as to attract the attention of passers-by to the writing on the placard – illustrates that the taking of action which draws the attention of a third party to the availability of matter in a manner which has the effect of enticing or encouraging the third party to take some step which results in that matter becoming available for his or her comprehension can be sufficient to amount to participation in publication of that matter. The question whether particular action amounts to enticement or encouragement of that nature is appropriately described as one of "fact and degree".

Thus, in Google Inc v Duffy with reference to Hird v Wood, the content of the snippet component of several Google search results was said to have been such as to "entice" a third party to click on the hyperlink in those search results to obtain more information. That was because the snippet component "naturally invite[d] the reader to click on the hyperlink for explanation and elaboration" of the matter, such as to be aptly described as "the electronic analogue of the person who places a post-it note on a book which reads 'go to page 56 to read interesting gossip about X'".

Although aspects of the reasoning in Duffy are capable of being interpreted more broadly, the outcome in Duffy is best understood as having turned on the particular content of the snippet component of the search result in that case. The outcome in Duffy cannot be generalised to the extent of indicating that the combination of elements of a search result will always, or even generally, operate to direct, entice or encourage a third party conducting a Google search to click on the hyperlink component for further information.

The search result in the present case was not of a kind described in Google Inc v Australian Competition and Consumer Commission as a "sponsored link", which is "a form of advertisement created by, or at the direction of, advertisers willing to pay Google for advertising text which directs users to a web site of the advertiser's choosing". The plurality in that case observed in respect of a sponsored link that "Google is not relevantly different from other intermediaries, such as newspaper publishers (whether in print or online) or broadcasters (whether radio, television or online), who publish ... the advertisements of others".

Like the search result in Duffy, the search result in the present case was rather of a kind described in Google Inc v Australian Competition and Consumer Commission as an "organic search result". That terminology was not used by the primary judge in the present case but can be taken as a shorthand description of the results that are generated from the operation of the ranking algorithm of the Google search engine, which her Honour described as using "various 'signals' or clues to identify what results the user is most likely looking for".

Unlike the position in Duffy, no feature of the content of the particular organic search result in the present case has been found to have operated as an enticement or encouragement to click on the hyperlink.

The conclusion of the primary judge that Google was a publisher, which was upheld by the Court of Appeal, was instead based on the broad proposition – necessarily applicable to the results of all Google searches – that the inclusion of a hyperlink in a search result is enough for the provision of the search result to amount to active and voluntary participation by Google in the process by which the matter referenced in the hyperlink is published to a user of the Google search engine. Her Honour said that "provision of a hyperlinked search result is instrumental to the communication of the content of the webpage to the user", that "[t]he Google search engine lends assistance to the publication of the content of a webpage on the user's device, by enabling the user to enter a search query and, a few clicks later, to view content that is relevant to the user's search", and that "[t]he inclusion of a hyperlink within a search result naturally invites the user to click on the link in order to reach the webpage referenced by the search result". I cannot accept a proposition of that breadth.

That Google's mission is "to organise the world's information and make it universally accessible and useful" by "connecting users to information on the internet that is relevant to their query and is of high quality", that it "has a commercial interest in providing a quality service with responsive search results", that the web crawler and indexing programs of its search engine are continuously in operation, and that the ranking algorithm of its search engine operates to identify and display those results ranked by relevance to identify what the searcher "is most likely looking for", do not alone or in combination affect what I consider to be the critical feature of an organic search result. The critical feature is that the search result is no more than a designedly helpful answer to a user-initiated inquiry as to the existence and location of information on the Internet.

By entering a search term into the Google search engine, the searcher looks for matter on a topic of interest to the searcher. By providing a search result, Google indicates where on the Internet that matter may be found. The hyperlink in the search result identifies the webpage on which matter on that topic is located. The hyperlink in the search result – no differently from any other hyperlink – also provides a shortcut which facilitates immediate access to the webpage should the searcher choose to take the further step of clicking on it. Having obtained the search result, including the hyperlink and the snippet, it is then up to the searcher to decide whether or not to take that further step of clicking on the hyperlink so as to access the webpage. Google does not, merely by providing the search result in a form which includes the hyperlink, direct, entice or encourage the searcher to click on the hyperlink.

I agree with the orders proposed by Kiefel CJ and Gleeson J.

KEANE J. I agree with Gordon J that Google's appeal should fail. In particular, I agree with her Honour that Google published the material defamatory of Mr Defteros ("the Underworld article"), as the courts below found. I also agree that the common law and statutory defences of innocent dissemination and qualified privilege are not available to Google in this case for the reasons given by the courts below. I write separately to explain my conclusion that Google published the Underworld article. Because mine is a dissenting judgment, I will be brief.

Google provided search results in response to users' search queries. Those search results directed its users to The Age's webpage containing the Underworld article. Through the hyperlinks provided by Google, users were thereby enabled to have direct and nearly instantaneous access to the Underworld article. All this occurred as Google intended by the operation of its search engine in accordance with its design and in the ordinary conduct of Google's business. Google thereby participated in the publication of the Underworld article to a user of its search engine for the purposes of the law of defamation in Australia. My conclusion in this regard is founded upon the primary judge's unchallenged findings of fact in relation to the design and operation of Google's search engine. To those findings I now turn. Google's search engine

Google's search engine is an automated information retrieval system which allows a user to navigate the vast amount of information available on the World Wide Web through user‑designed search queries or terms. Google states that its mission, in providing this service, is "to organise the world's information and make it universally accessible and useful" by connecting users to information on the Internet that is relevant to their query and is of high quality.

Each webpage of the World Wide Web has a unique Uniform Resource Locator ("URL"), which acts as the address for that webpage. Webpages are created, most commonly, with a code called HyperText Markup Language ("HTML"). The HTML code of a webpage is read by a software program called a browser, which displays text and images to a user on the user's device. A webpage typically contains hyperlinks, which are HTML code that contains a URL for another webpage. Clicking on a hyperlink will take a user from one webpage to that other webpage. Unless a user knows the precise URL for a webpage, he or she must find and access the webpage by some other means, most commonly, a search engine.

Google's search engine uses computer algorithms to make predictions about what webpages are most likely to be of interest to a user, based on the terms of the user's search query. Its search engine responds to a query from a user by first identifying what information is available on the Web, using a fully automated Web crawler program. That program determines which webpages to crawl and how often, as well as what type of information is collected; if a webpage is ranked as more important, it will be "crawled" more often.

Next, the Google search engine uses an indexing program to organise the data obtained by the Web crawler program into a form that is more easily searched by computer algorithms. The program creates an index, listing every webpage that contains each word found during the crawling stage, as well as other information such as whether the page comprises text, images or video and whether it appears to be a news article.

When a user enters a search query, the words from the query are evaluated by a series of algorithms against the information in the index, as it is at that point in time. A webpage's relevance is evaluated by reference to "signals" or clues as to what the user may be looking for, such as the number of times one or more of the user's search terms appears on the webpage, how recently the content of that webpage was published or updated, the location of the user, and the user's previous search history.

A search result is then presented – almost instantaneously – to the user comprising a list of webpages, ranked according to relevance as determined by the ranking algorithm. Generally, for each webpage, a user is shown the title of the webpage with the search terms in bold, a "snippet" of the content of the webpage and a shortened form of its URL. The title operates as a hyperlink, which the user can click on to be taken to the webpage.

Google's search engine also allows users to search for images, which are displayed to the user in search results as a series of "thumbnails". Each thumbnail contains a hyperlink to an interstitial page which provides more information about the image. Clicking on the interstitial page takes the user to the webpage whence the image originated.

Google's search engine, and its component programs, are both designed by humans. They operate as they are intended to do in accordance with that design. That is so, even though the component processes of a Google search are fully automated and completed without human intervention. Further, while the Google search engine is not capable of evaluating the meanings conveyed by the words or images displayed on a webpage, including whether they are true, false or defamatory, human intervention can, and does, occur at the point of Google's "legal removal" process. This process may alter the results that are shown to a user. In accordance with its various policies, Google may choose to remove a webpage, identified by its URL, from the search results that are returned by the Google search engine. This process does not remove the webpage from the Web; a user may still access it, for example, by using another search engine, or directly via the URL. The publication rule

A cause of action for defamation arises when a person suffers harm by way of damage to his or her reputation by the publication of defamatory matter to a third party. As was said by Kiefel CJ, Keane and Gleeson JJ in Fairfax Media Publications Pty Ltd v Voller, "[p]ublication is the actionable wrong".

It has long been the law in Australia that each publication to a third party is actionable as a separate tort. It has also long been the law that while publication must be intentional, a broad view has been taken of the intentional acts that will constitute publication. In Webb v Bloch Isaacs J said: "The term published is the proper and technical term to be used in the case of libel, without reference to the precise degree in which the defendant has been instrumental to such publication; since, if he [or she] has intentionally lent his [or her] assistance to its existence for the purpose of being published, his [or her] instrumentality is evidence to show a publication by him [or her]."

Similarly, in the words of Ribeiro PJ, in the Court of Final Appeal of Hong Kong in Oriental Press Group Ltd v Fevaworks Solutions Ltd: "A person was held liable for publishing a libel if by an act of any description, he [or she] could be said to have intentionally assisted in the process of conveying the words bearing the defamatory meaning to a third party, regardless of whether he [or she] knew that the article in question contained those words." This inclusive view of what amounts to publication was affirmed by this Court's decision in Voller.

In Crookes v Newton, the Supreme Court of Canada held that the provision of hyperlinks by the defendant in that case did not, without more, constitute publication by the defendant of the content on the webpage that appeared in the hyperlink, even if the hyperlink was followed and the defamatory content accessed. Abella J (with whom Binnie, LeBel, Charron, Rothstein and Cromwell JJ agreed) said: "A reference to other content is fundamentally different from other acts involved in publication. Referencing on its own does not involve exerting control over the content. Communicating something is very different from merely communicating that something exists or where it exists. The former involves dissemination of the content, and suggests control over both the content and whether the content will reach an audience at all, while the latter does not ... Hyperlinks are, in essence, references. By clicking on the link, readers are directed to other sources ... The ease with which the referenced content can be accessed does not change the fact that, by hyperlinking, an individual is referring the reader to other content."

The defendant in Crookes v Newton was not the operator of a search engine like Google; he owned a website, on which he posted an article with the impugned hyperlinks. Nevertheless, these observations by Abella J were cited to support Google's position in this Court. These observations were not directed to, and do not recognise, the nature and extent of Google's involvement in bringing a user of its search engine and a communication by another person (a "primary publisher") together.

The primary judge's findings establish that Google's search engine generates results ranked in a specific order through the use of Google's confidential and proprietary algorithms and methodology as the intended response to a query by Google's users. Success for Google in its business of operating its search engine consists of its users clicking on a hyperlink because they are satisfied with Google's response. This is as Google intends. To satisfy its users, Google ensures that its search engine is constantly learning from the large volumes of query data that it accumulates and processes: over 100 billion searches are made by its users every month, and of those more than 500 million each day have never been made before. A user's history of queries can provide useful information about his or her probable intentions.

In addition, Google's argument based on the reference to the passage from Crookes v Newton excerpted above pays little regard to the bilateral nature of publication, a consideration of significance in Australian law. The bilateral nature of publication

In Dow Jones & Co Inc v Gutnick, Gleeson CJ, McHugh, Gummow and Hayne JJ said: "Harm to reputation is done when a defamatory publication is comprehended by the reader, the listener, or the observer. Until then, no harm is done by it. This being so it would be wrong to treat publication as if it were a unilateral act on the part of the publisher alone. It is not. It is a bilateral act – in which the publisher makes it available and a third party has it available for his or her comprehension. The bilateral nature of publication underpins the long‑established common law rule that every communication of defamatory matter founds a separate cause of action."

In Voller, Kiefel CJ, Keane and Gleeson JJ adopted this statement of principle as a correct statement of the law relating to publication for the purposes of the law of defamation. So did Gageler and Gordon JJ, who went on to say: "Publication of matter by means of the Internet is accordingly complete when and where the matter is accessed by a third party in a comprehensible form."

In this Court, it was argued on Google's behalf that a person whose participation in the process of communicating defamatory matter to a third party consists of assistance to the reader, rather than to the primary publisher, is not relevantly a publisher. It was said that to characterise the activity of a person who aids a reader in having access to the defamatory writing as participation in publication of the writing would be an absurd extension of prima facie liability for defamation. In light of Dow Jones v Gutnick, it cannot sensibly be denied that a person who aids another to comprehend defamatory matter does participate in the publication of that matter to that person. There is nothing novel in this: a person who reads a defamatory writing to a blind or illiterate person publishes that writing, just as a person who reads a newspaper article aloud over the radio publishes the article.

It has never been suggested in the authorities that a defendant in such a situation has not published a statement to a third party because that statement was originally published by another potential defendant. Nor has it been suggested in the authorities that the circumstance that a defendant has published a defamatory statement to a third party at the third party's request precludes a finding of publication by the defendant. Whether the publication of a statement defamatory of a third party has been made in response to a request for information may be relevant to whether the publication has occurred on an occasion of qualified privilege, but that this is so only confirms that, being a circumstance relevant to a matter of defence, it is not such as to negative the element of publication essential to the existence of prima facie liability for defamation.

To characterise Google's role in the publication of the Underworld article to a user of its search engine as assistance to the user which was, therefore, not assistance to The Age as the primary publisher involves several errors. The first is the fallacy of the excluded middle, the logical error that assistance to the user cannot be of assistance to The Age as well. The Age's interest as a news media publication is, after all, to reach a wide readership, and, like Google, it is indifferent as to the identity of those readers.

The second error in this aspect of Google's argument lies in its downplaying, or indeed denial, of the significance of the circumstances in which Google brings its users and The Age together. The publication of defamatory material, which occurred when a user of Google's search engine gained access to the Underworld article, occurred by reason of the assistance intentionally provided by Google in the course of its business. That publication would not have occurred but for Google's facilitation (save for the rare case where a person may directly access a webpage by its URL). No doubt, to say this is not sufficient reason to provide an affirmative answer to the question whether Google intentionally facilitated the publication. But to say that Google was simply "assisting the publishee", as counsel for Google submitted, is to fail to appreciate the nature and extent of Google's participation in that exercise. The design of Google's search engine includes the presentation of search results by way of hyperlinks which are apt to, and which are meant to, afford near‑instantaneous access to the one or more webpages which Google indicates answer a user's query. The ease and immediacy of that access are integral to Google's business. That differentiates Google, both from the owner of a telephone network which passively conveys statements being made by others via its cable lines, and from a publisher on the Internet who inserts hyperlinks into his or her own content as references to another text.

This case is not like Crookes v Newton, where the defendant's hyperlinks were in the nature of footnotes to his article which, so far as his work was concerned, readers might or might not click on, that being a matter of indifference to him. It is not a matter of indifference to Google that the user chooses to click on a hyperlink. That the user should make that choice is the essence of Google's business, which, as it says itself, is "to organise the world's information and make it universally accessible and useful" by connecting users to information on the Internet that is relevant to their query.

Google's search engine cannot be accurately described as a passive instrument by means of which primary publishers convey information. Google's business consists of the automated provision of relevant responses to requests for information and the enabling, by the use of hyperlinks, of near‑instant access to that information at the choice of the user. Google actively ranks those responses by relevance – that is, as was said in argument on Google's behalf, in an algorithmically attempted understanding of the searcher's interest. It may be acknowledged that Google does not contribute to the content of the works which its search engine disseminates, just as it may be acknowledged that Google's search engine facilitates access to those works by opening the way to the primary publisher's webpage; but neither of these facts is inconsistent with the conclusion that Google has intentionally brought the work of the primary publisher to the attention of the user.

Thirdly, while Google and The Age have not come together to agree to expand the publication of The Age's work, Google and The Age each acted with the intention of bringing the Underworld article to a readership of third parties. The symbiotic relationship between Google and The Age has not been reduced to an agreement, but each party pursues its intention as an unremarkable part of its ordinary business. The intention of putting The Age's articles before an audience is manifested by Google's search engine responding to requests from its users for access to information supplied by primary publishers such as The Age. In this regard, it operates as it is intended by its designers and owner to do. That Google's users make use of its search engine to gain access to information supplied by others is in no way inconsistent with the intention that is common to both Google and The Age; indeed, it is the consummation of that intention. And that is so no less because the webpage containing the defamatory matter is The Age's webpage rather than Google's, or because access to that webpage depends on the choice of Google's user. The benefit of the Internet

The plurality in Crookes v Newton expressed a concern that an unduly broad view of what is involved in publication might restrict "the flow of information and, as a result, freedom of expression", thereby creating a "chill" that risks "impairing [the] whole functioning" of the Internet. But new technologies are not launched into a world free of the laws that provide redress for injury to others; and the social value of innovation is not absolute. The concern that the availability of redress to all injured by the provider of a new technology should not unduly interfere with society's enjoyment of its benefits has traditionally been addressed, in the context of defamation, in the balancing exercises involved in the recognised defences. It has not previously been thought necessary to modify the broad scope of the traditional publication rule to render harm inflicted by new technologies entirely immune from redress. That would be the case if Google's argument were to be accepted. In Voller, Gageler and Gordon JJ said: "Formulated in the age of print, the common law rule was not seen to require modification in order to deal with the advent of the telegraph or the telephone. When, in the late 19th century, an operator employed by a telegraph company at one place transmitted a customer's message to an operator employed by the same telegraph company at another place who then reduced the message to writing and delivered the message in print to the addressee, the telegraph company was as much a publisher of the printed message as was the customer. When, in the late 20th century, one person communicated with another by telephone, the telephone company was not a publisher. The difference was that 'a telegram [was] sent only through the direct participation of agents of the telegraph company whereas in the case of a modern‑day telephone call the caller communicate[d] directly with the listener over the facilities of the telephone company, with no publication by the company itself'."

To the objection that the broad understanding of publication for the purposes of the law of defamation adopted in Australia extends potential liability for defamation too widely, one may respond further that the breadth of potential liability has always been a feature of the approach affirmed in Webb v Bloch. An expansive publication rule is warranted to ensure that all persons injured by a defamatory publication should have a remedy against each person responsible for inflicting that injury whatever the "precise degree" of his or her instrumentality may be. And, as noted above, the concern that the scope of actual liability for the injury should not be unduly wide has been addressed through defences such as innocent dissemination, qualified privilege and fair comment. In the application of these defences, issues as to the relative social utility of the publication as against harm to individual reputation are addressed and balanced so as to ameliorate the potential for injustice of the broad approach to publication. Conclusion

The findings of the primary judge establish that Google's search engine operates, as intended in accordance with its design, to respond to a user's search query by facilitating near‑instantaneous access by hyperlink to publications on topics relevant to the user's query. For the purposes of the law of defamation in Australia, that is sufficient communication of the content of the work of the primary publisher to the user of Google's search engine. This is no less true because the publication appears on the webpage of the primary publisher rather than on Google's webpage, or because the user's access to the publication depends on his or her choice to take advantage, as Google intends, of the assistance rendered by Google's hyperlink. Whether Google is liable for the injury caused by that publication will depend on the application of one or more of the defences available to a publisher in its position.

Google's appeal should be dismissed with costs.

GORDON J. In early 2016, the respondent, Mr Defteros, became aware that using the "Google search engine system" to search the term "george defteros" produced search results displayed on www.google.com.au which included: "Underworld loses valued friend at court -SpecialsGanglandKillings ... www.theage.com.au > Features > Crime & Corruption ▼ June 18 2004 - Pub bouncer-turned-criminal lawyer George Defteros always prided himself on being able to avoid a king hit – The Age Online" The first line, "Underworld loses valued friend at court ‑SpecialsGanglandKillings ...", was a hyperlink to an article published in The Age Online at www.theage.com.au ("the Underworld Article"). As the second line of the Google search result recorded, the Underworld Article appeared in the "Features" section of The Age Online under the heading "Crime & Corruption". The third and fourth lines of the search result are what is described as a "snippet", or a summary, of the Underworld Article. By clicking on the first line of the result – the hyperlink – the reader was immediately taken to the Underworld Article published by The Age Online at www.theage.com.au.

Mr Defteros sued the appellant, Google LLC, for defamation. Mr Defteros did not allege that the search result itself defamed him, but that the Underworld Article did. He said that, by its publication of the search result containing the hyperlink, Google published the Underworld Article. Google denied that it was the publisher of the Underworld Article and, relevantly, pleaded defences of common law and statutory innocent dissemination, and common law and statutory qualified privilege. There was no dispute in this Court that the Underworld Article was defamatory.

The Google search engine system is in fact a number of systems – including the web crawler program, the indexing program and the ranking algorithm. And although the Google search engine system operates in a "fully automated" manner, the systems of which it is comprised are designed by humans and operate as they are intended to operate. In general terms, the web crawler and indexing programs operate without being triggered by a search query. The web crawler and indexing programs are designed to determine, among other things, the relevant "importance" of certain webpages that they crawl and index. The web crawler program ranks the webpages it crawls, determining which webpages are "important" so as to crawl those more often. The indexing program notes the characteristics of the webpages crawled by the crawler program, including whether a webpage "appears to be a news article". The ranking algorithm uses various signals or clues to determine the order in which to display results to users. Those clues include another important rank called PageRank, which evaluates how often other webpages link to that webpage, and the "importance" of those linking webpages. Creation and continuous operation of that system is the antithesis of passivity.

Google published the Underworld Article. To conclude otherwise is contrary to the strict publication rule, recently restated by this Court in Fairfax Media Publications Pty Ltd v Voller. Google intended to publish the Underworld Article in the sense that its conduct was active and voluntary. Google intentionally participated in, lent its assistance to, was instrumental in and contributed to the communication of the Underworld Article by identifying, indexing, ranking and hyperlinking it within the search result. It matters not that Google was unaware of the alleged defamatory content of the Underworld Article. The Underworld Article was accessed in a comprehensible form by third parties who clicked on the hyperlink that the Google search engine system provided in the search result. The fact that the third parties had to click on the hyperlink does not alter the conclusion that Google published the Underworld Article. For a third party to access defamatory material in a comprehensible form often, if not always, involves an act by that third party – the turning of the head; the selection, buying and then opening of a newspaper and turning the pages of the newspaper; or, here, entering search terms and clicking on a hyperlink.

Google, as the creator and operator of the Google search engine system, and The Age Online, as the initial publisher of the Underworld Article, also have an objective common intention. Although an objective common intention is not necessary to amount to publication, it will be sufficient. An objective common intention is found in this case – to facilitate access to news articles – by The Age Online publishing the Underworld Article at www.theage.com.au and by Google's tailoring of its search engine system – in particular the web crawler and indexing programs – to specifically identify, locate and index news articles published on the web, to rank the results returned in response to a particular user's search query and then to tailor the display of those results for that user.

And that Google tailors its search engine system in that way is unsurprising. It is Google's business model – it "has a commercial interest in providing a quality service with responsive search results". That is why it pays particular attention to webpages that appear to be news articles. It is why the Google search engine system crawls for and separately indexes news articles. News articles are of particular importance to Google. Google and online news providers such as The Age Online (which generate the news articles) have an objective common intention – to facilitate access to news articles. It is why Google participates in the publication of news articles to which its search engine system provides a hyperlink.

When it was said in argument that Google was "agnostic" as to what hyperlinks a user will click on, that expression not only obscured the objectives, elements and operation of the Google search engine system, but was inconsistent with them. To adopt and adapt the language in Voller, Google's attempt to portray itself as passive has an air of unreality. Having taken action to obtain a commercial benefit by creating and operating a search engine that facilitates access to news articles, it cannot deny that it is involved in the publication of those news articles.

Innocent dissemination is a defence to a cause of action in defamation. It is not a denial of the element of publication. The defence, however, cannot be made out if the disseminator knew of the defamatory character of the publication, or that the matter was likely to be defamatory. Here, Google was fixed with knowledge that Mr Defteros claimed the material was defamatory when a solicitor employed by a firm of which Mr Defteros is the principal lodged a removal request on Google's website for the Underworld Article to be removed from Google's search results. Google was provided with the Uniform Resource Locator ("URL"). The removal request form was provided and generated by Google. Although some of the information provided in Mr Defteros' removal request was inaccurate, that is not determinative. Reasons of principle, practicality and fairness dictate that a plaintiff is not required to give a defendant notice of any particular imputations which they claim are conveyed or the basis upon which any defences would fail. Google was therefore aware of the defamatory character of the Underworld Article a reasonable time after having been given notice and the defence of innocent dissemination cannot be established.

At common law, an occasion of qualified privilege arises where there is a reciprocity or community of duty or interest between the recipient and the publisher. Mere curiosity or a matter's mere quality as news is not sufficient. And an interest is generally not found if the publication is to the world at large. There was no such interest here; not only was there no evidence of such interest but there were uncontested findings of fact that a subset of the third parties clicked the hyperlink out of idle curiosity.

There was no dispute that the statutory defence of qualified privilege is wider than the common law defence. But the essential element remains – that the recipient has a relevant interest. Where, as here, a small number of persons had clicked on the link out of mere curiosity, the statutory test could not and cannot be met in respect of those persons.

To explain those conclusions, it will be convenient to proceed by examining: first, the Google search engine system and Google's business model; second, the strict publication rule; and third, the objective common intention of Google and The Age Online. Against those considerations, it will then be necessary to look in some detail at particular cases and examples given significance in argument, including the "graffitied wall" cases and the decision of the Supreme Court of Canada in Crookes v Newton. Finally, it will be necessary to say something more about the defences relied upon by Google. The Google search engine system

How the Google search engine system operates lies at the heart of this appeal. The relevant factual findings of the trial judge were not disputed in the Court of Appeal or in this Court. Those findings were consistent with earlier decisions that have considered the Google search engine system. It is to those findings I now turn.

The Google search engine system and its component programs are designed by humans who work for Google. When a user enters a search query into the Google search engine system, algorithms which employees of Google have developed, but which are fully automated, make predictions about which webpages are of interest to that user and display search results as a list of hyperlinked text. But the results – the end product – are only part of the story. It is necessary to understand how the underlying programs that comprise essential aspects of the Google search engine system were intended to and do operate.

First, there is the "web crawler" program, which is "fully automated" and "uses a large number of computers to constantly visit and process webpages on the [w]eb". That program determines which webpages to crawl, how often, and what information is to be collected from each webpage. The web crawler program has its own ranking system, and deems certain webpages to be "important", which "are crawled for new data more frequently than less important pages", and the data of re‑crawled pages is updated. The web crawler program brings this data about the webpages back to Google's servers.

Second, an "indexing program" then "organise[s] the data obtained by the [w]eb crawler into a form that is more easily searched by computer algorithms". It "builds a list", which is constantly refreshed by the results of the web crawler program, of every webpage that contains "each word" found during the crawling stage, and notes other aspects, including each webpage's publication date, whether the webpage comprises text, images or video "and whether the webpage appears to be a news article". The index contains "each word and a list of the unique IDs that relate to the webpages that contain that word".

Third, there is a "series of algorithms". "[W]hen a user enters a search query", the algorithms evaluate the words from that query "against the information in the index, as it is at that precise point in time". A "ranking algorithm" uses various "'signals' or clues" to identify what result the user "is most likely looking for" and displays those results to the user as a list of webpages, "ranked according to relevance, as estimated by the ranking algorithm". Those signals or clues include: (1) the number of times one or more of the user's search terms appear on the webpage, as indexed by the indexing program; (2) how often other webpages link to that webpage, and the importance of the linking webpages (this is the signal known as PageRank); (3) how recently the content of that webpage was published or updated (freshness); (4) the order in which the search terms appear on that webpage; (5) the location of the user, as determined from the user's Internet Protocol or IP address; and (6) the user's previous search history.

The results displayed to a user are typically listed as the titles of webpages, which are hyperlinked; "snippets" of the content of the webpage; and a shortened form of the webpage's URL. The user's search terms are shown in bold. There may be many pages of search results, and the first page contains those results which the ranking algorithm has determined are most relevant to the search query.

As has been mentioned and as is readily apparent, the web crawler and indexing programs are not reliant for their operation on, or triggered by, a particular user's search terms. It is the algorithms, especially the ranking algorithm, which are triggered by the user's search query. It is the stored and constantly updated data crawled and indexed by the web crawler and indexing programs against which the algorithms then evaluate the search terms. That is, the web crawler is always crawling and the indexing program is always indexing, and when a search occurs, the words of that search are then evaluated "against the information in the index, as it is at that precise point in time". Contrary to Google's submission, its vast repository of information obtained and organised by the web crawler and indexing programs is not "an undifferentiated mass until a search is requested". And in crawling and indexing, news articles are a particular, if not primary, focus. A webpage which appears to be a news article, importantly, is separately crawled – identified – and indexed as such a page. "Important" webpages are crawled more often. The web crawler program and the ranking algorithm's focus on "important" webpages, the crawling of such pages for updated data more frequently, and the PageRank and freshness clues used by the ranking algorithm then combine to produce search results in response to a search query.

The design of the Google search engine system as a whole is intended to, and does, affect the results that are produced when a user enters a particular search query. And it is for that reason that, in seeking protection for its search results under the First Amendment to the Constitution of the United States, Google has itself successfully argued that the design of its search engine system involves the exercise of evaluative judgment.

Although the operation of the Google search engine system is automated, Google can manually affect search results. By human intervention, Google can alter search results by removing a webpage from its search results, but such removal does not remove the page from the web. Removal of a webpage from search results is done in limited circumstances including "legal removal". This is a process that can be triggered by a user lodging with Google a removal request, a form on Google's website which asks a user to explain why they believe that the content complained about is unlawful. This was the process used in this case. It will be necessary to return to consider it later in these reasons. Google's business model

The way the Google search engine system operates can only be fully appreciated in light of Google's business model. Before the trial judge in this case, the Vice President of Engineering at Google gave evidence that Google's mission is "to organise the world's information and make it universally accessible and useful". The trial judge found that Google sought to do this "by connecting users to information on the internet that is relevant to their query and is of high quality". That is consistent with the statement in the Full Court decision in Google Inc v Duffy that: "[t]he very purpose of a search engine like Google is to find webpages on the World Wide Web which contain the information for which users are searching. For that very reason those users are likely to follow one or more of the hyperlinks in the displayed paragraphs to the webpage. Human curiosity is a powerful force which in large part, explains the exponential growth of the [World] Wide Web. The ease of reference provided by internet search engines entices many to 'surf' the internet."

Google "has a commercial interest in providing a quality service with responsive search results": more users and a more tailored search and listing of third-party content means more valuable advertising opportunities. That is why Google pays particular attention to webpages that appear to be news articles. Through its interactions with news outlets, Google benefits because its service is more "responsive" and of better "quality". News outlets also benefit from what is, in effect, a referral service provided by the Google search engine system.

That the relationship between Google and news media outlets is a two-way relationship was given statutory recognition by the enactment in 2021 of the Treasury Laws Amendment (News Media and Digital Platforms Mandatory Bargaining Code) Act 2021 (Cth), enacted after the Australian Competition and Consumer Commission's 2019 inquiry into digital platforms. One aspect of the legislative amendments was the introduction into the Competition and Consumer Act 2010 (Cth) of a "News media and digital platforms mandatory bargaining code". Under that scheme, the responsible Minister may make a declaration that a service is a "designated digital platform service". "If the designated digital platform service makes available covered news content of [a] registered news business" (emphasis added), the responsible digital platform corporation must, among other things, provide certain information to the news business and, if the digital platform corporation is notified by the news business that the news business "wishes to bargain over one or more specified issues relating to the registered news business' covered news content made available by the designated digital platform service", "[e]ach bargaining party must negotiate in good faith". For the purposes of the relevant Part of the Competition and Consumer Act, "a service makes content available if ... a link to the content is provided on the service" (emphasis added), and in such a case "a user of a service interacts with content made available by the service if ... the user interacts with the link" (emphasis added). Although the Minister has to date not designated any digital service platforms, after the amendments to the Competition and Consumer Act, and "[c]onsistent with the policy objectives of the Code", Google "reached agreements with a range of ... news businesses".

The relevance of this is two-fold. First, the commercial benefit that Google obtains by creating and operating its search engine, particularly in relation to news outlets, is inconsistent with its assertion of passivity and confirms the conclusion that its acts of participation amount to publication for the purpose of the strict publication rule. Second, it underlines the objective common intention of Google and news outlets, which is sufficient but not necessary to establish that Google is a publisher. Strict publication rule

The strict publication rule was recently restated by this Court in Voller. Accordingly, after Voller, the rule can be summarised as follows: first, any person who, by an intentional – in the sense of active and voluntary – act, participates, assists or is instrumental in, or contributes to any extent to the process directed to making defamatory matter available for comprehension by a third party is a publisher. All degrees of such participation amount to publication. Such participation may involve acts which precede the comprehension by third parties of the defamatory material and, in all cases, a person is a publisher regardless of whether they knew that the matter contained defamatory content. Finally, publication is a bilateral act which occurs when the matter is available to be comprehended, and is accessed in a comprehensible form, by a third party.

For an act to be "intentional" it is sufficient that it is "active and voluntary". It is not necessary for a publisher to have knowledge of the defamatory content. And not every active and voluntary act which ultimately results in a third party accessing defamatory content is a publication. There must be an active and voluntary participation in the process that is in fact directed to making matter available for comprehension by a third party. If a person writes a defamatory statement which is placed in a locked drawer only for a thief to break in and publish it, the writer is not a publisher of the statement. The writer has not engaged in a process directed to making matter available for comprehension by a third party.

An important feature of the strict publication rule is that acts which precede the comprehension by third parties of defamatory material can amount to publication, and that such acts may include providing a platform for the communication of defamatory matter. The principle that acts which precede the comprehension by third parties of the defamatory material can amount to publication acknowledges that publication is a process directed to making matter available for comprehension by a third party. For example, the clapping down of a printing press was held to amount to publication, and, in the case of Voller, so was the contracting with Facebook for the provision of a Facebook page, and posting content on that page which gave Facebook users the ability to comment on that content. Put in different terms, the breadth of activity captured by the strict publication rule is vast.

And importantly, liability as a publisher for any degree of participation in the process of publication, including acts which precede the comprehension by third parties of the defamatory material, is dependent upon the bilateral acts of a defendant making matter available for comprehension, coupled with its actual comprehension. In the context of the internet, the act of making matter available was described in Dow Jones & Co Inc v Gutnick as "uploading", whereby a person places the material on a webpage, managed by a web server, from which a person wishing to have access to that material may access it in a comprehensible form by issuing a request to the relevant server in the form of nominating the webpage's URL. That is, the fact that publication requires the act of a third party does not mean that the person making that matter available is not a publisher.

Participation in a process of making matter available to be comprehended by a third party may take many forms. In addition to composing the defamatory material itself, it will include, for example, directing someone to compose material for the purpose of publication even if there was no direction that the material be defamatory, physically pointing to defamatory material so as to draw attention to it, and providing a platform for the dissemination of defamatory material, either by providing a website which hosts a discussion forum, or by posting an article on Facebook upon which persons can comment. Common to all of these examples is that a defendant has provided some part of the means by which a third party may, by their own acts, comprehend the defamatory material. That is, the publisher's acts form part of the process directed to making matter available for comprehension.

Acts forming part of such a process might afford a publisher a defence of innocent dissemination, but, as Voller makes clear, those acts will still amount to publication under the strict publication rule. Consistently with that principle, the statutory defence expressly assumes that a person may be a publisher even if the person: "(a) was not the first or primary distributor of the matter; and (b) was not the author or originator of the matter; and (c) did not have any capacity to exercise editorial control over the content of the matter (or over the publication of the matter) before it was first published." And, in Australia, the strict publication rule has not been modified to apply differently to new technologies. The question whether it should be modified to address the Google search engine system is addressed later in these reasons.

To adopt and adapt the language used in Voller, Google intentionally created and operates the Google search engine system – a platform – which crawls the web, indexes the material on the web and then, in response to an inquiry from a user, interrogates the indexed materials, ranks the materials identified as responding to the request, specifically identifies the results containing news, and then provides the user with not only the ranked list of results but the hyperlink to and snippets of the news articles.

Where, as here, the creator and operator of a search engine system has the intention that third parties will access and read news articles that are hyperlinked in the search results generated by that system, it is a publisher of the news articles under the strict publication rule. That is this case. Objective common intention

Prior to Voller, the concept of joint tortfeasors was thought to be as applicable to the law of defamation as it is to all other torts. To be liable as a joint tortfeasor, the principle required that there be "a concurrence in the act or acts causing damage". Mere or knowing assistance was not sufficient – there had to be "an objective common intention". The view that concurrence, assent or approbation is required to establish publication was observed in Voller to be inconsistent with the strict publication rule. Under the strict publication rule, all degrees of participation in the process of publication amount to publication; "mere communication" of the defamatory matter to a third person is enough. That is not to say that common intention is irrelevant. Having a common intention to publish, under joint tortfeasor principles, will be sufficient for a defendant to be characterised as a publisher; such an intention is not necessary.

In this case, Google and The Age Online do have an objective common intention. As the explanations of Google's business model and the operation of the Google search engine system make clear, Google's assertion in argument that it is "completely agnostic as to what hyperlinks a user will click on" must be rejected. Google and The Age Online have each engaged in "concerted actions towards a common end" – to communicate the Underworld Article to readers. That the Google search engine system might be characterised as "assisting" its users to access news articles does not negate that it has an objective common intention with The Age Online – to facilitate access to news articles to those users. Graffitied wall cases

Earlier cases – referred to as the "graffitied wall" cases – have concerned whether an owner of a building or noticeboard on which a defamatory statement was initially affixed by a third party was a publisher because they had "consented to, or approved of, or adopted, or promoted, or in some way ratified, the continued presence of" the defamatory statement so that other persons could continue to read it. Google called in aid such concepts, submitting that it should not be characterised as doing any of those things when the Google search engine system (which it created and operates) displays its hyperlinked search results. As was explained in Voller, "[t]here is plainly a critical difference between that line of cases, involving defendants who have played no role in the facilitation of publication prior to becoming aware of the defamatory matter" (emphasis in original) and cases such as the present.

As has been explained, and as with the media companies in Voller, the operation of the Google search engine system which yielded the search result hyperlinked to the Underworld Article was an intentional act of facilitation by Google which preceded both the comprehension by third parties of the Underworld Article, and Google's knowledge of the defamatory nature of that article. Relevance of Crookes v Newton

A lot of emphasis was given in argument to the decision of Abella J in Crookes, in which Abella J concluded that "a hyperlink, by itself, should never be seen as 'publication' of the content to which it refers". That emphasis is misplaced for two reasons. First, Abella J relied in reasoning to that conclusion on two lines of authority – first-instance decisions of the United Kingdom and decisions from North America. Both lines of authority are inconsistent with the strict publication rule. Second, Abella J's treatment of hyperlinks as "references" which do not communicate the content of that to which they refer and require an act of a third party to comprehend that content is inconsistent with the application of the strict publication rule to publication by reference. It is necessary to address each of these matters in turn. First-instance decisions of the United Kingdom

Abella J made express reference to the decisions of Eady J in Bunt v Tilley and Metropolitan International Schools Ltd v Designtechnica Corpn, which she said suggested that "some acts are so passive that they should not be held to be publication", and noted that in Bunt it was held that there must be "knowing involvement in the process of publication of the relevant words" (emphasis in Bunt). Bunt and Metropolitan International Schools are inconsistent with the strict publication rule. Under the strict publication rule, "that a person merely plays a passive instrumental role in the process" of publication does not preclude them from being a publisher because "all degrees of participation in publication are publication". As has been explained, Google intends to and does participate in the publication. And under the strict publication rule, there is no requirement for "knowing involvement in the process of publication of the relevant words" (emphasis in original); a person is a publisher regardless of whether they knew that the matter contained defamatory content.

Google cannot, in any event, be described as playing a "passive instrumental role in the process" of publication. Google is not the company that makes the router that connects the user's computer to the internet; Google is not the internet service provider that connects the user to the web; and, contrary to Google's submission in argument, its position is in no way analogous to a person who supplies fuel to a truck that delivers a newspaper which contains defamatory material. North American authorities

The United States authorities referred to by Abella J, which concerned the single publication rule, also appear to have been a particular influence on Abella J's reasoning. The Australian law of defamation differs markedly from the path that has been taken in North America. That difference was identified and emphasised in Dow Jones. The core difference is that Australia treats every publication as a separate cause of action, whereas the prevailing rule in the United States, originating in the courts of New York, has been a single publication rule. Under that rule, a plaintiff can only bring one action for all publications made of defamatory material, unless it can be established that a subsequent publication falls within the "republication" exception to that rule, discussed below. Decided within a jurisdiction where the single publication rule was firmly rooted and in the context of the republication exception, the approach taken in the New York cases of Klein v Biben and MacFadden v Anthony, endorsed by Abella J, was that a bare reference to an earlier publication does not give rise to a new or separate cause of action. The single publication rule was squarely considered and rejected by this Court in Dow Jones.

Moreover, the United States cases, and a Canadian case referred to by Abella J which applied them, were not concerned with hyperlinks. Klein was concerned with whether a statement in a newspaper article that further details could be found in an earlier article constituted republication of the earlier article. MacFadden was concerned with whether a radio broadcast "calling attention to" a magazine article constituted republication of the article. Carter v BC Federation of Foster Parents Assn, which applied Klein and MacFadden, was concerned with, among other things, whether a reference in a newsletter to an online forum constituted publication of comments on the forum. In each case, the answer given was "no". Those decisions are inconsistent with the strict publication rule as restated in Voller and, in particular, the application of that rule to publication by reference, which is considered below.

The "simple hyperlink" with which Abella J was concerned is, in any event, not a sufficient description of the Google search engine system and the search results generated by it. Abella J considered that "[a] reference to other content is fundamentally different from other acts involved in publication". Abella J went on to state that: "[r]eferencing on its own does not involve exerting control over the content. Communicating something is very different from merely communicating that something exists or where it exists. The former involves dissemination of the content, and suggests control over both the content and whether the content will reach an audience at all, while the latter does not." (emphasis in original)

Not only is that distinction inconsistent with the strict publication rule, but the context – the reason why Abella J drew that distinction – is important. Abella J referred to the "ancillary" nature of the participation of a person making reference to defamatory material, even where a person's goal is "to expand that publication's audience", and observed that in such circumstances "the allegedly defamatory information has already been made available to the public by the initial publisher or publishers' acts". That led Abella J to distinguish hyperlinks "from acts in the publication process like creating or posting the defamatory publication, and from repetition". Such reasoning exhibits the influence of the republication exception to the single publication rule. The rationale for that exception is that the subsequent publication "is intended to and actually reaches a new audience", and it does not apply where an audience is referred to a defamatory publication's "preexisting dissemination". As has been explained, the single publication rule provided the context for the approach taken in Klein and MacFadden, upon which Abella J relied, but it forms no part of the law in Australia.

Although Abella J correctly identified that the hyperlink gives the hyperlinker – described in the reasons as the primary author, a concept not known in the law in Australia – no control over the content in the secondary article to which they have linked, Abella J also correctly identified that "the person selecting the content to which [they] want[] to link might facilitate the transfer of information" (emphasis in original), which is "a traditional hallmark of publication". That is the strict publication rule. Publication by reference

It is then necessary to address the fact that Abella J treated a hyperlink as a "reference" and considered that hyperlinks "share the same relationship with the content to which they refer as do references". As Abella J put it, "[b]oth communicate that something exists, but do not, by themselves, communicate its content". The analogy is inapt and unhelpful. First, a defendant may be liable for publication by referring to a statement originally published on another occasion by them or another. The latest edition of Gatley on Libel and Slander, under the heading "Publication by reference", puts it in these terms: "It is clear that a defendant may be liable for publication by referring to a statement originally published on another occasion by himself or another: eg if A writes a defamatory publication of C and D then writes, 'A description of C may be found in A's work'. Whether a defendant is liable on this basis will be a matter of fact and degree. In such cases it may be that the correct analysis is not so much that D republishes what A has written but that he procures a publication of it by leading readers to it, for what they see is the original work of A, though that makes no difference to D's liability. An obvious modern instance would be where the defendant incorporates on a website a link to another site containing material defamatory of the claimant." (emphasis in original)

Among the cases cited in Gatley is Buchanan v Jennings, a decision of the Privy Council on appeal from the Court of Appeal of New Zealand. A member of the Parliament of New Zealand made a defamatory statement about the plaintiff in the House of Representatives. That was protected by absolute privilege. He later told a journalist, outside the House, that he "did not resile" from his earlier statement inside the House. The plaintiff sued, relying on the statement to the journalist as a republication by reference of the earlier statement. The question was whether the absolute privilege protecting the statement made in the House protected the statement made outside the House. The Board held that it did not. In reasoning to that conclusion, Lord Bingham said: "It is clear that at common law every republication of a libel is a new libel and a new cause of action. The republisher of the libel may or may not be the same as the original publisher. The republication may or may not be made on an occasion enjoying any privilege (whether absolute or qualified) attaching to an earlier publication or republication. It is further clear (see Gatley on Libel and Slander, 10th ed (2004), para 6.33) that a defendant may be liable for republishing by reference to a statement originally published on another occasion by himself or another." (emphasis added)

In 2014, the Court of Appeal of England and Wales in Makudi v Baron Triesman of Tottenham held that a "reference to an earlier parliamentary statement" is capable of being "taken as a fresh publication of it" and there is "no significance to the distinction ... between repetition and reference". Similarly, in the Canadian decision of Spike v Golding, a newspaper had published an article which made reference to an earlier defamatory article it had published after proceedings had been instituted by the plaintiff in respect of the first article. The second article referred to those proceedings and stated that the publisher "was prepared to prove all the paper had stated and a good deal more". The Nova Scotia Supreme Court held that "publication was sufficiently shown". And a reference may be made not only by the original publisher, but by another party, and yet constitute publication.

Second, Abella J said that hyperlinks and references both require some act on the part of a third party before they gain access to the content. An act of a third party – comprehension by a third party – is a necessary element of publication, as publication is bilateral. The fact that that element is addressed separately is not new. A third party will often, if not always, have to act to access defamatory information – to turn their head to see a defamatory placard, to turn the pages after reading a headline on the front page of a newspaper or a chapter heading in the table of contents of a book, to scroll through their Facebook feed, or, as here, to click on a hyperlink.

But the two matters relied upon by Abella J – that a hyperlink is a mere reference and that to access the content requires an act of a third party – also demonstrate that observing that a hyperlink by itself is "content-neutral" is itself a neutral consideration; it is not determinative. Knowledge of defamatory content has never been a necessary element for publication. Knowledge is addressed in the available defences. Does the strict publication rule require modification?

Before turning to the defences, it is necessary to revisit a question addressed in Voller: "Formulated in the age of print", does the strict publication rule require modification in order to deal with search engines and hyperlinks or, more particularly, the Google search engine system and the hyperlink in this case? The law of defamation has consistently had to grapple with technological advances, which are "much older than the Internet and the World Wide Web". The common law was not seen to require modification in order to deal with the advent of the telegraph or the telephone, radio or television, the internet or social media. And it has not been shown to require modification in order to deal with the Google search engine system and the hyperlink in this case. Google was a disseminator of the Underworld Article – a publisher. An innocent disseminator is still a publisher but has a defence to a cause of action in defamation, not a denial of the element of publication. Innocent dissemination

The principles are not in dispute. Innocent dissemination is a defence to liability as a publisher, not a denial of the element of publication, developed by courts to mitigate the harshness of the strict publication rule. To succeed in a defence of innocent dissemination at common law or under the Defamation Act 2005 (Vic), the onus, relevantly, is on a subordinate distributor who claims the benefit of the defence to prove that they did not know that the publication contained a libel; that they did not know, and had no ground for supposing, that the publication was likely to contain a libel; and that their ignorance was not due to any negligence on their part.

The present appeal concerns the second element, which Gaudron J in Thompson v Australian Capital Television Pty Ltd said "can conveniently be expressed as whether the distributor knew or ought to have known that [the publication] contained a libel". As expressed in the Defamation Act, the question is whether the subordinate distributor "neither knew, nor ought reasonably to have known, that the matter was defamatory", and that that lack of knowledge "was not due to any negligence on the part of" the subordinate distributor.

The issue that divided the parties was precisely what knowledge was required for the common law defence and s 32(1)(b) of the Defamation Act. Google submitted that something other than a "mere notification" of a claim should be required, and that, "[a]t the very least", there should be a notice setting out "the imputations of concern and ... an explanation as to why [those imputations] cannot be justified or excused". In support of that contention, Google referred to the decision of Lord Denning MR in Goldsmith v Sperrings Ltd, in which his Lordship (in dissent) held that the defence should apply unless the subordinate distributor "knew or ought to have known that the [publication] contained a libel on the plaintiff ... which could not be justified or excused". Google sought to have this Court focus on the quality of the notice said to be necessary to defeat its lack of knowledge for the purposes of its defence. For reasons of principle, practicality and fairness, Google's submission should be rejected.

At a level of principle, the defence of innocent dissemination is one of "confession" and "avoidance". That is, once publication has been established, as it has been here, the onus is on the defendant to satisfy a court of its lack of knowledge or constructive knowledge of the defamatory nature of the material. Acceptance of Google's submission would require that the plaintiff prove that the notice given to the subordinate distributor was sufficient to negate its presumed lack of knowledge. That is contrary to principle because it inverts the onus of proof.

Further, the requirement that the subordinate distributor lack knowledge of the defamatory nature of the material is not concerned with whether the material is actually defamatory, in the sense that it is actionable. The bar has been and remains lower. In Emmens v Pottle, Lord Esher MR and Bowen LJ each said that the defence will not be available if the subordinate distributor knew or ought to have known that the material was "likely" to contain a libel. Similarly, Romer LJ in Vizetelly v Mudie's Select Library Ltd said that there must not have been anything in the work "which ought to have led him to suppose that it contained a libel" (emphasis added). Or, as Gaudron J said in Thompson, all that is required is that the subordinate distributor "neither knew nor ought to have known of the defamatory character" of that material. Such statements must be understood so as only to require, for the defence to fail, that the subordinate distributor knew or reasonably ought to have known, for any reason, that the material was "likely to lead an ordinary reasonable person to think less of the person concerned". What will be sufficient will vary from case to case and from libel to libel.

That conclusion is reinforced by matters of practicality and fairness. Google's submission, put differently, is that a potential plaintiff should only be able to defeat the defence of innocent dissemination if the plaintiff details the imputations sought to be conveyed by the allegedly defamatory content and explains why there are no available defences. To simply state that proposition is to see how impractical and unfair such a position would be to the vast majority of potential plaintiffs. The trial judge and the Court of Appeal correctly rejected Google's submission in this case. The notice given to Google fixed it with knowledge that Mr Defteros claimed that the Underworld Article was defamatory. As has been explained, a solicitor completed a removal request form provided on the Google website and included the URL to the Underworld Article, which Google's employee could have clicked. The fact that inaccurate statements were made to Google in that removal request and subsequent correspondence about the Underworld Article does not and cannot alter the position that, a reasonable period after the completion of the removal request form, Google was aware of the claimed "defamatory character" of the article and could read the article. In those circumstances, after a reasonable time, which on the evidence was one week after the removal request, Google could not establish its defence of innocent dissemination. Qualified privilege Common law

At common law, a defamatory statement receives qualified protection when it is made in discharge of a duty or in furtherance or protection of an interest where the recipient of the defamatory statement has a corresponding duty to receive or interest in receiving it. Reciprocity of duty or interest is essential. As was stated in Bashford v Information Australia (Newsletters) Pty Ltd, such an "occasion" may exist where a statement is: "fairly made by a person in the discharge of some public or private duty, whether legal or moral, or in the conduct of his own affairs, in matters where his interest is concerned. ... If fairly warranted by any reasonable occasion or exigency, and honestly made, such communications are protected for the common convenience and welfare of society".

The concept of "interest" is used "in the broadest popular sense, as when we say that a [person] is 'interested' in knowing a fact – not interested in it as a matter of gossip or curiosity, but as a matter of substance apart from its mere quality as news". A corollary of that proposition is that an interest is generally not found if the publication is to the world at large because "a person has no legitimate interest or duty" in making such a publication merely because "the subject matter of the publication is one in which the public is interested". And the emphasis is on "duties and interests rather than the state of mind of the defendant". In asking the ultimate question whether an occasion is privileged because there is such reciprocity of duty or interest between the maker and the recipient of the matter, it is to be asked, "[w]hat legal, social, or moral duties or interests were engaged between the [defendant] as publisher and those [persons] to whom it published"?

In determining whether the occasion was privileged, the court examines all the circumstances of the case including "the nature of the defamatory communication, the status or position of the publisher, the number of recipients and the nature of any interest they had in receiving it, and the time, place and manner of, and reason for, the publication". It is after considering these matters that the court "makes a judgment as to whether the publisher had a duty or interest that justified making the publication and whether the recipients, or some of them, had a duty to receive or interest in receiving" the material. This usually involves questions of public policy.

Google submitted that the "common convenience and welfare of society as a whole" would be served by finding that Google has "an interest or duty to publish search results" hyperlinking to matter that is responsive to a user's search terms. In so submitting, it relied on the fact that the trial judge had held, under the statutory defence, that a significant number of the people to whom Google published the Underworld Article had a "legitimate interest[]" in reading it, including because they were "looking for a lawyer" or were "considering employment with [Mr Defteros'] firm". Google contended that requiring it to remove its results where it cannot prove that all users have a sufficient interest would subvert "the greater interest ... to the lesser interest", which "does not serve the common convenience and welfare of society as a whole".

Google bore the onus of establishing the defence. The concurrent findings in the courts below were that Google had failed to establish that it provided its service to its users as a matter of legal, social or moral duty and failed to establish that it had a community or reciprocity of interest with those users.

That is unsurprising. Here, Google published the defamatory material to users of its search engine system, including a small number of persons who accessed it out of "idle interest or curiosity". The extent of the publication is relevant in determining whether the occasion is privileged. Publication to users without a legitimate interest is not privileged. As Higgins J made plain in Howe v Lees, it is not sufficient that the interest in the published material arises out of curiosity or because of the material's "mere quality as news". It is publication to a particular person that is protected.

Moreover, because the focus is on the community of duty or interest of the publisher and the recipient, contrary to Google's submissions, the phrase "common convenience and welfare of society as a whole" records a result and explains why the communication and the relevant duty or interest gave rise to an occasion of qualified privilege. The phrase "common convenience and welfare of society" comes from the judgment of Parke B in Toogood v Spyring, which was endorsed by this Court in Lange v Australian Broadcasting Corporation and Bashford. Although in dissent in the result, McHugh J in Bashford explained, in terms which are not inconsistent with those of the plurality in that case and which were approved in Papaconstuntinos v Holmes à Court, that such concepts do not determine whether the occasion is privileged. "They must be distinguished from the question whether society would recognise a duty or interest in the publisher making, and the recipient receiving, the communication in question" (emphasis added) and evidence must be adduced to show that "both the givers and the receivers of the defamatory information had a special and reciprocal interest in its subject matter" (emphasis in original). In this case, the trial judge had no such evidence, and her Honour and the Court of Appeal were therefore correct to hold that no duty or interest of the requisite kind existed. Statutory qualified privilege – s 30 of the Defamation Act

At trial and before the Court of Appeal, Google was successful in establishing its defence of statutory qualified privilege in relation to a "substantial proportion" of the 150 persons to whom the trial judge found it had published the Underworld Article. Its appeal to this Court in relation to the statutory defence is therefore confined to that "small number" of persons who the trial judge found had clicked on the hyperlink to the Underworld Article "out of idle interest or curiosity", which her Honour assessed for the purpose of damages as 50 persons.

Google submitted that the Court of Appeal, and the Full Court in Duffy, erred in proceeding on the basis that, for the purposes of s 30 of the Defamation Act, Google had to establish that its users had a "legitimate" interest in matter published to them by Google. Google submitted that the statutory defence is wider than the common law defence and extends to "any matter of genuine interest or 'apparent' interest". It submitted that, on the facts of this case, it was sufficient that: users had sought information about "george defteros" and had clicked on the hyperlink to the Underworld Article; the topic of the Underworld Article was of "considerable public interest"; the Underworld Article was "published by a reputable news source"; and a Google employee decided not to remove the Underworld Article from the search results in accordance with Google's "Reputable Source Defamation Push Back Policy", which policy the trial judge found was reasonable. That submission is rejected. It ignores the express words of s 30(1)(a) of the Defamation Act – namely, that "the recipient has an interest or apparent interest" (emphasis added). That interest must be a "legitimate interest", in the sense that it is "a matter of substance apart from its mere quality as news". And the trial judge's finding that Google's "Reputable Source Defamation Push Back Policy" and its employee's reliance on it were reasonable does not alter the fact that Google's belief as to such an "apparent interest" could not be reasonable in respect of persons who did not hold that requisite interest.

In support of its contention that the statutory defence extends to "any matter of genuine interest", Google referred to a number of decisions of the New South Wales Court of Appeal and the decision of the Privy Council in Austin v Mirror Newspapers Ltd. The most appropriate summary of those cases can be found in the judgment of Hodgson JA (with whom Mason P and McColl JA agreed) in Echo Publications Pty Ltd v Tucker [No 3]. Adopting that summary, Hodgson JA later explained that the high-water mark of the cases is that the sense of "interest" in the (former) New South Wales statutory defence of qualified privilege is "substantially wider" than that required under the common law.

Even if that be accepted, as his Honour's survey of the authorities demonstrates, matters of "curiosity" or matters of interest for their "mere quality as news" are still not captured by the statutory defence. Accordingly, as no challenge was, or is, made to the trial judge's finding that a small number (assessed for the purpose of damages as 50 persons) clicked the link out of such curiosity, the defence cannot be made out in relation to those persons. Conclusion

The appeal should be dismissed with costs.

EDELMAN AND STEWARD JJ. The respondent is a solicitor who specialises in the field of criminal law. In 2004, a Melbourne newspaper, The Age, published an article ("the Underworld article") that was said to contain certain defamatory imputations concerning the respondent. The appellant is a multinational technology company that either itself or through its subsidiaries operates the Google internet search engine. In early 2016, entering the respondent's name into that search engine (or, to use the vernacular, "googling it") produced a series of results, one of which was a hyperlink to the Underworld article.

In February 2016, a solicitor employed by Defteros Lawyers (the respondent's law firm) completed a Google "removal request form" in respect of the Underworld article. The appellant did not remove the article from the results generated by its search engine. Thereafter, about 150 individuals clicked the hyperlinked search result through to the Underworld article after googling the respondent's name. The issues for determination are whether the appellant was a publisher of the Underworld article for the purposes of the tort of defamation, and, if so, whether the defences of innocent dissemination and qualified privilege were available to it. For the reasons that follow, the appellant did not publish the Underworld article. The facts in amplification

The primary judge found that the Underworld article conveyed a defamatory imputation, namely that the respondent had crossed the line from being a professional solicitor to being a confidant and friend of criminal elements. That finding was not challenged. It is nonetheless noteworthy that, whilst the respondent was successful below, a consequence of the proceeding has been the reproduction of the Underworld article in the reasons of the primary judge. Anyone now using the Google search engine to find those reasons will naturally also find that article, albeit in a different place.

The learned primary judge set out in detail the operation of the internet and of the Google search engine. Her Honour observed that each webpage has a Uniform Resource Locator ("URL") and that the code commonly used for creating webpages is called HyperText Markup Language or HTML. If one knows the URL of a webpage, one can type that URL into the address bar of a web browser and be taken to that webpage. Absent such knowledge, internet search engines – such as Google – provide a critical means of navigating the internet. Entering keywords into a search engine produces relevant search results. At the very least, those search results will contain hyperlinks to the URLs of potentially relevant and useful webpages, the vast majority of which are published by third parties. Sometimes the search results might contain additional words or phrases taken from a given webpage: this additional information is called a snippet.

The Google search engine is highly successful and very popular with users of the internet. It is based upon web crawling and indexing to collate, rank, and organise the data from trillions of webpages. It then applies sophisticated algorithms to produce, within usually half a second, search results from those trillions of webpages. The algorithms yield results that are not displayed randomly but are ranked based upon an assessment of the relevance to the searched item. Every month, over 100 billion searches take place on Google. The Google search engine is fully automated; no human intervention is needed. However, humans designed the search engine and invented the algorithms it uses; thus, search engines generally operate "precisely as intended by those who own them".

It is in the appellant's interest to have a popular search engine. That is because the appellant or its subsidiaries, or both, carry on an advertising business. Whilst not the subject of specific findings below, using the Google search engine produces at least two different types of search results. The first set of results are advertisements that relate to the subject matter of the search; the second are the search results themselves. A different algorithm is used to produce the advertisements. These proceedings were not concerned with this aspect of the appellant's global business and different considerations may apply to advertisements with respect to the issue of publication.

Googling the respondent's name in 2016 produced, amongst other things, the following result: "Underworld loses valued friend at court -SpecialsGanglandKillings ... www.theage.com.au > Features > Crime & Corruption ▼ June 18 2004 - Pub bouncer‑turned‑criminal lawyer George Defteros always prided himself on being able to avoid a king hit – The Age Online"

In 2016, clicking on that hyperlinked search result produced the Underworld article. The respondent accepted that no part of this search result otherwise involved the publication of any defamatory material.

The respondent did not sue The Age for defamation; instead, in 2010, the respondent sued two authors of a book that, amongst other things, contained a chapter based on the Underworld article. One of the two authors was also the author of the Underworld article. The claim settled at mediation, and it was a term of the settlement that the authors would make certain revisions to the book. In exchange, the respondent released the authors from all liability in relation to a number of matters, including any article published in The Age or its affiliated publications concerning the respondent.

On 4 February 2016, the solicitor at Defteros Lawyers completed the Google "removal request form". He provided the URL of the Underworld article and made the following representations: "In 2007 the subject of this article, Mr George Defteros, sued the publisher in defamation in the Victorian (Australia) Law Courts. The article was found to be defamatory and the publisher settled the matter, paying a confidential settlement sum. It was a term of the settlement that the article be removed from the internet."

Each matter represented above was false. The respondent had never sued the publisher of the article; the article was never found by a court to be defamatory; the publisher never settled any proceeding with the respondent and never paid him any settlement sum; no term of any settlement existed that required the article to be removed from the internet; and no settlement of any proceeding with the publisher has ever existed.

An employee of the appellant (or of one of its subsidiaries) who worked in the Google legal removals team responded to the removal request by asking for a copy of the court order. The solicitor responded with more falsehoods. He said that the matter had been settled in a mediation before it proceeded to trial and that the terms of settlement were confidential, but that the publisher had conceded that the article was defamatory and had agreed to remove the article from its website and accordingly from the internet. Again, none of that was true.

In 2016, the appellant had a "Reputable Source Defamation Push Back Policy". That policy listed news sources that the appellant considered reputable, including The Age. In accordance with that policy, the appellant's employee sent the solicitor from Defteros Lawyers an email informing him that the appellant had decided not to remove the "content" from its search engine. The email stated that the appellant did not control the content of third‑party webpages and encouraged the solicitor to deal directly with The Age. The primary judge decided that it was reasonable for the appellant to rely on sources it knew to be reputable in order to determine what content should be removed. It was also reasonable for the appellant to refer the solicitor to The Age given that the solicitor had contended that a settlement had previously been entered into between the respondent and that newspaper. Finally, given the business environment in which the appellant operates, the primary judge also found that it was not unreasonable for it to take no further steps to determine the accuracy of the Underworld article. None of these findings was challenged on appeal.

Between 11 February 2016 and 24 December 2016, about 150 people clicked on the hyperlink to the Underworld article following a Google search of the respondent's name. The primary judge found that a substantial proportion of those who undertook the searches had a legitimate interest in receiving information about the respondent, for the purposes of the statutory defence of qualified privilege. For example, some people were looking for a lawyer, were existing clients looking for the firm's contact details, or were considering employment with Defteros Lawyers. However, a small number had no such legitimate interest and searched the respondent's name out of idle interest or curiosity. The case below

It was never suggested that the appellant was a publisher of the Underworld article – which was the work of a third party – merely upon the presentation on a computer screen, or like device, of the search results themselves, although no one disputed that the appellant was a publisher of those very results.

Instead, applying the decision of this Court in Webb v Bloch, the primary judge found that the appellant became the publisher of the Underworld article each time a person clicked on the hyperlink to that newspaper article. In Webb, Isaacs J said: "The term published is the proper and technical term to be used in the case of libel, without reference to the precise degree in which the defendant has been instrumental to such publication; since, if he has intentionally lent his assistance to its existence for the purpose of being published, his instrumentality is evidence to show a publication by him."

Echoing the foregoing language, the primary judge decided that because the provision of the hyperlinked search result was "instrumental" to the communication of the Underworld article and because this lent "assistance" to the publication of that article to a searcher of the respondent's name, the appellant published the Underworld article. As the primary judge observed, "[t]his is exactly what the Google search engine is designed to do". In her Honour's view: "the provision of a hyperlink within a search result facilitates the communication of the contents of the linked webpage to such a substantial degree that it amounts to publication of the webpage".

The primary judge was also of the view that the appellant was only a secondary and not a primary publisher. The respondent did not dispute this. Given that the appellant's search engine must traverse trillions of webpages on billions of occasions each month, inferentially the appellant will have no actual knowledge of the content of those pages, save in exceptional circumstances. It will also, inferentially, and subject to exceptional circumstances, know nothing about: the authors of those webpages; why the webpages have been created; and whether the information they convey is accurate. Again, inferentially, and subject to exceptions, the appellant will know few specifics about the users of its search engine or their motivations for making particular searches.

Because of the foregoing, according to her Honour, the appellant will "almost always, if not always" have a defence of innocent dissemination at common law before it is given notice that its search engine has produced a link to defamatory material. But the primary judge considered that once a proper notification has been given, the appellant would then have a reasonable period of time to remove the webpage from its search results before it may be liable. The evidence suggested that a reasonable period of time was one week.

Here, and despite its falsity, the primary judge was satisfied that the content of the removal request of 4 February 2016 sufficiently notified the appellant that its search results were producing a link to the Underworld article, which the respondent claimed was defamatory of him. It followed that, from 11 February 2016, the defence of innocent dissemination was not available to the appellant for what was said to be its publication of the Underworld article during 2016 to about 150 individuals who had googled the name of the respondent. However, a defence of statutory qualified privilege was made out in relation to a substantial proportion of these acts of publication. The primary judge awarded general damages, after mitigation, of $40,000.

The reasoning of the primary judge is consistent with the conclusion of Beach J in Trkulja v Google Inc LLC [No 5], where his Honour decided that it was open to a jury to conclude that the appellant was a publisher of third‑party webpages hyperlinked in results generated by using its search engine. The appellant's argument in that case – that it could not have been a publisher because of the passive nature of its involvement – was rejected. Beach J said: "To say as a general principle that if an entity's role is a passive one then it cannot be a publisher, would cut across principles which have formed the basis for liability in the newsagent/library type cases and also in those cases where someone with power to remove a defamatory publication chooses not to do so in circumstances where an inference of consent can be drawn."

A similar conclusion was reached by McDonald J in Trkulja v Google Inc. On appeal, this Court observed that McDonald J was correct to conclude that it was "strongly arguable" that the appellant's intentional participation in the communication of allegedly defamatory results to search engine users supported a finding of publication.

The Court of Appeal of the Supreme Court of Victoria upheld the primary judge's decision in this case. It referred extensively to the decision of the Full Court of the Supreme Court of South Australia in Google Inc v Duffy. The Court of Appeal decided that the relevant issue for determination might be framed – as it was in Duffy – as follows: "whether [in relation to the Underworld article] Google's role as facilitator through its search engine [was] sufficiently proximate to the display of the search results [and the text of the Underworld article when its hyperlink was clicked on] [so as] to constitute participation in the publication of [the Underworld article]".

In Duffy, the Full Court decided that the appellant was the publisher of hyperlinks and snippets that were themselves defamatory and that more directly drew the defamatory content of the third‑party webpage to the attention of the searcher. The Court of Appeal in this case recognised that Kourakis CJ had left open the possibility that the appellant might be a publisher even if the hyperlink or snippet did not itself contain defamatory material. Kourakis CJ had observed that there may be circumstances, "depending on both the ease with which the hyperlink can be accessed, and the information provided by the hyperlink (whether or not that information is defamatory in itself)", in which a search result might incorporate and thereby publish a third‑party webpage.

The Court of Appeal was also influenced by the judgment of Hinton J in Duffy. His Honour was of the view that a snippet "entices" a searcher to click upon a link to a webpage. Without needing to resort to concepts of adoption or endorsement of the kind considered in Byrne v Deane, Hinton J held that "the hyperlink bespeaks a willingness on the appellant's part to transport the enticed searcher immediately to the relevant web page for more information – to publish the web page to those who, having read the snippet, want more information". The act of enticement was an act of publication.

The Court of Appeal agreed with Hinton J's analysis, even though the search results in this case did not contain any snippets featuring defamatory material. The Court reasoned as follows: "We also agree with the observation of Hinton J in Duffy (FC) that a search result (snippet) may entice a searcher to click on the hyperlink to obtain the additional information available in response to the searcher's request. In the present case, the Underworld article Search Result (containing the hyperlink to the Underworld article) was an enticement to the reader, in the way referred to by Hinton J, to click on the hyperlink to obtain more information about Mr Defteros when it published the words 'Underworld loses valued friend at court', 'SpecialsGanglandKillings', 'Crime & Corruption' and 'Pub bouncer‑turned‑criminal lawyer George Defteros always prided himself on being able to avoid a king hit – The Age Online'." (footnote omitted) Routes to a conclusion of publication

There are four ways in which a defendant might be found to be a publisher of a communication to a person that is defamatory of a different person. The first is where a defendant performs the actual act of communication. The other three are based on the attribution to a defendant of another's acts that form part of the process of communication: (ii) where a defendant authorises another to perform an act of communication and the communication is within the scope of that authority; (iii) where a defendant assists, in the sense of procures, provokes or conduces, another in performing the act of communication; and (iv) where a defendant ratifies or adopts the communication of another.

Putting to one side cases of negligent publication, the negligence for which, on one view, might better be treated as governed by principles developed in the law of negligence, a defendant will only be responsible as a publisher if the defendant intended to perform the act of communication (instance (i)) or had a common intention, in the sense of an intention to communicate that is shared with the person whose communication they were authorising, assisting, or ratifying (instances (ii), (iii) and (iv)). Again putting negligence aside, it has been established for more than a century that in every instance a person must have an intention to communicate before they can be a publisher.

The concept of intention in the tort of defamation is not a unique idea dreamed up by ingenious defamation lawyers. It has exactly the same meaning as the intention in other torts of strict liability such as trespass to land, trespass to goods, or false imprisonment. To "intend something is to intervene in the world to bring about a chosen consequence". In the law of defamation, the relevant intention is generally proved by showing acts that reveal an "objective or manifested intention" to communicate. It is an intention that the act of communication be performed or an intention to authorise, assist, or ratify the act of communication being performed by another; it is not concerned with the consequences of the act or with issues of fault, such as the reasonableness of the act.

The simplest illustration of the meaning of intention for the tort of defamation is the example of Lord Esher MR, cited on many occasions, of a person who writes a defamatory letter and places it in a locked drawer which is opened by a thief who communicates the contents of the letter to the world. There has been unanimity for more than a century that the writer is not a publisher. But what is the difference between that circumstance of non-publication and the circumstance of publication where the person writes exactly the same letter but leaves it in a prominent place where it is read by a third party? Both people have performed a voluntary act. In both cases the voluntary act was central to the communication of the defamatory material. But the difference is that the acts of the person who left the letter in a prominent place are sufficient for an inference that the person had "an intent to publish"; they had performed "an act for the purpose of communicating [the matter] to a third person".

The method of constituting a person as a publisher based on a common intention (sometimes described as a common design) is not confined to defamation but is applicable to all joint tortfeasors. In that respect, the decision of this Court in Lee v Wilson does not support any stricter and exceptional rule for the tort of defamation. In that case, Dixon J said that liability depended upon "mere communication of the defamatory matter to a third person". But when his Honour in the following sentence went on to say that "communication may be quite unintentional", he was not excluding a requirement of an intention to publish. Read in context, this was a reference to a communication regardless of knowledge as to whether what had been conveyed was defamatory. Absent the need for such knowledge, the transmission of information is but a "mere communication".

The decision of this Court in Fairfax Media Publications Pty Ltd v Voller concerned the posting of potentially defamatory statements by third parties as comments on news stories posted on the Facebook pages of certain media outlets. The issue for determination was whether the media outlets were also the subsidiary publishers of those third‑party comments. This Court divided on the relevance of, and answer to the question of, whether the appellants intended the communication of all comments made by users on the appellants' Facebook pages.

Kiefel CJ, Keane and Gleeson JJ said that the "intention of the author of the defamatory matter is not relevant because the actionable wrong is the publication". However, their Honours later considered the circumstance of a person who "intentionally lent [their] assistance" to the existence of defamatory matter for the purpose of publication and described that circumstance as "apposite" to the finding by Isaacs J in Webb of the liability of the solicitor enlisted by the defendants who had composed the defamatory circular "for the purpose of publication" and subsequently "consciously distributed it". Their Honours added that the finding "points to the fact that it is the defendant's act of participation in publication which must be intentional, in the sense of being voluntary".

Also in the majority, Gageler and Gordon JJ held that "all degrees of intentional participation in the process of publication constitute publication for the purposes of the law of defamation". Their Honours concluded that the appellants had posted material on their Facebook pages "with the intention that third parties [would] comment on the material posted".

We dissented in Voller on the application of the concept of intention to the facts. In our view, by having on their Facebook pages the functionality for third-party "comment" upon stories posted by the appellants, the appellants could not possibly have intended to communicate, or to assist in the communication of, any and every remark by third parties upon their Facebook pages. All that they intended to communicate, or assist in the communication of, were those remarks that could genuinely be described as "comments" on the posted stories. The only route to publication in this case No actual communication, no authorisation, and no ratification

The Underworld article was not written by any employee or agent of the appellant; it was written by a reporter with no connection to the appellant, and published by an independent newspaper over which the appellant had no control or influence. The appellant did not in any way authorise the article to be written or published. And there was no submission that the appellant had ratified the communication of the article. Indeed, it was not suggested that the appellant had any prior knowledge of the existence of the article until February 2016, some 11 years after it first appeared in The Age. Even then, and following notification, the primary judge found that the appellant had acted reasonably in not removing the article from its search engine, and encouraging the solicitor from Defteros Lawyers to deal directly with The Age. The only real argument that might establish the appellant as a joint tortfeasor was that the appellant assisted The Age in publishing the Underworld article by providing the search results pursuant to a common intention shared with The Age to publish the Underworld article. Assistance with a common design or a common intention to publish

In Thompson v Australian Capital Television Pty Ltd, Brennan CJ, Dawson and Toohey JJ said of joint tortfeasors: "As was said in The 'Koursk', for there to be joint tortfeasors 'there must be a concurrence in the act or acts causing damage, not merely a coincidence of separate acts which by their conjoined effect cause damage'. Principal and agent may be joint tortfeasors where the agent commits a tort on behalf of the principal, as master and servant may be where the servant commits a tort in the course of employment. ... Otherwise, to constitute joint tortfeasors two or more persons must act in concert in committing the tort. Torts of all kinds may be joint and defamation is no exception." (emphasis added, footnote omitted)

In other words, the requirement for a "common" intention is a requirement for the intention to be concurrent or shared and not merely a coincidence of separate intentions. Hence, in order to act with a common intention there must be a "concerted action towards a common end", which will exist in the case of a subsidiary publisher where that person has "assisted" the primary publisher "with a common intention to publish".

The necessary degree of assistance may be minor, but "mere assistance is not sufficient" of itself to render a person a subsidiary publisher and a joint tortfeasor. The person must be a party to a common intention of some kind. The authorities thus distinguish between merely facilitating (or merely assisting) the doing of an act and assisting the doing of an act pursuant to a common intention. As Stuart‑Smith LJ said in Credit Lyonnais Bank Nederland NV (now Generale Bank Nederland NV) v Export Credit Guarantee Department: "it seems to me to be well established that a person who acts with another to commit a tort in furtherance of a common design will be liable as a joint tortfeasor. It is not enough that [they] merely facilitate[] the commission of the tort unless [their] assistance is given in pursuance and furtherance of the common design." Assistance in a common design, or with a common intention, is neatly expressed in many cases as a requirement that, as Steward J put it in Voller, the joint tortfeasor has "procured, provoked or conduced" another in the commission of a tort.

Here, in order for the appellant to be responsible as a joint tortfeasor with The Age, the intention to publish must be one that is common as between The Age and the appellant, because The Age is the primary publisher. It is not a common design or a common intention as between the appellant and each individual who googled the respondent's name. Those individuals are not publishers and they are not tortfeasors.

A conclusion that a person has (i) authorised, (ii) procured, provoked, or conduced, or (iii) adopted or ratified the act of publication must include a finding that the person has a common intention to publish. A person will not be a publisher without such intention. In some cases, a person may facilitate the act of publication of defamatory material but have no intention to convey that material. In such a case the person is not a publisher. As Edelman J said in Voller: "A bookstore owner or a newspaper vendor will generally be a publisher because, having chosen the book or newspaper to be sold and having made the sale, they have manifested an intention to make the communication to a third party even if they are not aware of its contents. On the other hand, the need for an objective intention to communicate to a third party means that a cashier in a bookstore will not be a publisher. Nor will a mere courier or postal worker who delivers a defamatory publication, or a person who mistakenly 'delivers one paper instead of another'. This principle also explains why telephone companies and internet service providers who passively transmit have been held not to be publishers. And it explains why the passive deliverer of a spare newspaper to a librarian, who 'never intended to publish', was held not to be a publisher, unlike '[a] printer and publisher [who] intends to publish'. In all of these instances of passive assistance, the basis for the conclusion is that there is no manifested intention to communicate any content." (footnotes omitted)

In addition, a defendant's reference to a communication in circumstances where they have no control over the content of the communication may sometimes be inconsistent with, and negate, a finding of a common intention or design to publish. Thus, in Crookes v Newton, the addition of hyperlinks to other webpages on Mr Newton's website did not make him the publisher of the content of those other webpages. The hyperlinks were characterised by the Supreme Court of Canada as mere references to material that Mr Newton did not control, and thus did not publish. Abella J, delivering the judgment of Binnie, LeBel, Charron, Rothstein and Cromwell JJ and herself, said: "A reference to other content is fundamentally different from other acts involved in publication. Referencing on its own does not involve exerting control over the content. Communicating something is very different from merely communicating that something exists or where it exists. The former involves dissemination of the content, and suggests control over both the content and whether the content will reach an audience at all, while the latter does not." (emphasis in original)

Whether it is correct to describe the results produced by the appellant's search engine as merely "references" may be doubted. On the one hand, there is an analogy between an application of the algorithm designed by the appellant to produce search results and the intellectual activity of Mr Newton in posting an article on his website that included hyperlinks to other webpages. On the other hand, Mr Newton was not the owner of a multinational business that deployed a search engine, based upon web crawling, indexing, and the application of algorithms, to expose its users to relevant search results and tailored advertisements. An essential assumption or general expectation of that business must be that, ordinarily, one or more of the hyperlinked search results will be clicked upon, thus causing the communication of material from a third‑party webpage. Whether that assumption or general expectation is sufficient to make the appellant a publisher of those third‑party webpages is addressed below. Publication of the Underworld article

In 2016, about 150 individuals googled the respondent's name. The appellant's search engine then produced a series of search results, one of which included a hyperlink to, and snippet of, the Underworld article. This search result, including the snippet, did not contain any defamatory material. This case is unlike Duffy. It was submitted by the respondent, nonetheless, that the appellant in 2016 was a publisher of the Underworld article in three distinct ways: the systems employed by the appellant; enticement; and incorporation. Each was broadly based upon a submission that the appellant assisted The Age with a common intention to publish the Underworld article, although the third basis conflated different routes to publication. The first basis: the systems employed by the appellant

The first contention was that the appellant was the publisher of the Underworld article because of the systems it employs – its web crawler program, indexing program, and ranking algorithm – which were instrumental to the communication of the defamatory matter. This contention adopted the reasons of Beach J and McDonald J in the Trkulja cases, set out above, and was said by this Court in the Trkulja appeal to be "strongly arguable". It was also the view of the primary judge in this case. This approach rejected any suggestion that the appellant performed a merely passive role when its search engine was used to produce results that included the hyperlink to the Underworld article. The appellant was the great designer; it invented the search engine, it maintained the search engine, and, through its servers, it enabled the search engine to operate. All of these matters were vital to the appellant's business; the appellant needed its search engine to be used to maximise its users' exposure to the advertisements that appeared on its website.

It may be accepted that "but for" the appellant's search engine, the defamatory matters in the Underworld article would not have been conveyed to the 150 individuals who viewed it via the Google search results. When the hyperlink to the Underworld article appeared, the search engine was operating in precisely the way its designers intended. But the appellant's role in that conveyance rose no higher than that of mere assistance or facilitation. When the appellant provided a list of search results and snippets through the processes described above, it had no common intention shared with The Age for the publication of the Underworld article any more than it had a common intention shared with those responsible for webpages listed in the many other search results and snippets.

The critical step that results in publication is that of the person searching and clicking on the chosen hyperlink. The role of the appellant rose no higher than a mere facilitator because the appellant had no common intention shared with The Age that the searcher click on the hyperlink to the Underworld article. In that respect, a clear distinction must be maintained between the act of publishing the selection of search results and snippets, which the appellant does, and the act of conveying material on third‑party webpages.

At best, the appellant's business assumes that ordinarily one or more webpages listed in the search results may be visited, but that is a matter entirely reserved for the searcher. Once the results have been published, it is the searcher who decides whether to click on one or more of the hyperlinks. In that respect, it was accepted on the facts of this case that the appellant could not be a publisher of the Underworld article unless and until this took place. It follows that the appellant in no way participated in the vital step of publication without which there could be no communication of defamatory material – namely the searcher's decision to click on the hyperlink of a particular result.

This appeal therefore does not present the occasion to consider whether the conclusion would be different in respect of those hyperlinks that, by agreement with a third party, are promoted by the appellant following a search request. Nor was any issue raised on this appeal about any service provided in the aggregation of news results. It suffices to say that it is arguable that the appellant and a third party might share a common intention to publish the content of a third‑party webpage that, as a consequence of an agreement between the appellant and the third party, is promoted as a search result.

Of course, the mere fact that a person to whom the communication is made must do some act to receive the defamatory material – even when that act is beyond the control of a putative publisher – does not always preclude a conclusion that the putative publisher assisted the primary publisher with a common design or a common intention to publish. In Vizetelly v Mudie's Select Library Ltd, the owners of a "circulating library" circulated or lent out a book called "Emin Pasha: his Life and Work". Volume 1 of this work referred to Mr Vizetelly being "not yet sober". A third party needed to attend the library and select this book for the defamatory comment to be read. The owners of the library were sued by Mr Vizetelly. They said that when they circulated this work, they did not know that it contained defamatory material as their library was too big for them to have read everything in it. They did not employ readers to check the content of the books lent out "because it was cheaper for them to run the risk, i.e., of publishing libels and being sued for those libels". The owners were found to be publishers of the book. The owners shared a common intention with the primary publishers of the book to disseminate the book and thus communicate its defamatory contents.

The Google search engine is distinctly unlike the library in Vizetelly. The appellant does not acquire a finite number of webpages, or some part thereof, and then make them available to be searched. The appellant does not own or control the internet. Unlike the readers who might have been employed in Vizetelly to check for defamatory content, it was not suggested that the appellant has any sensible means of being able to vet the content of every website in the world. And unlike the person who borrowed the book about Emin Pasha, the user of the Google search engine is, from the perspective of the appellant, an entirely unknown, unpredictable, and uncontrollable person searching the internet, which contains trillions of unknowable, unpredictable, and uncontrollable webpages.

If any analogy might be drawn, and again putting to one side the appellant's promotion of various advertised links (which is not in issue in this case), the analogy would be to position the appellant as an extremely knowledgeable guide working for profit in an enormous library, who facilitates a reader's search by providing the reader with a number of very specific pieces of information about where to locate a selection of items in the library and some indication of what they might contain. Importantly, the analogy is entirely inapt unless the appellant is a guide who is not employed by the library, is not a contractor to the library, and shares no interest with the library. The appellant has no common intention with the library or the publishers of books contained in it.

Acceptance of the fact that the appellant carries on a business that encourages individuals to use its website to conduct internet searches does not compel any contrary conclusion. The nature of that business justifies a conclusion that there is an expectation or assumption that one or more hyperlinks in the search results will probably be engaged when a search takes place. But this expectation or assumption does not suffice for an inference that the appellant's act of communicating a link to a third‑party webpage was done with a common design or a common intention shared with the third party.

The appellant is thus not the publisher of those third‑party webpages even though the use of its search engine was important – in a causative sense – to their ultimate publication. The first basis for publication is rejected. The second basis: enticement

The second way in which it was said that the appellant was the publisher of the Underworld article was that a search result entices a searcher to click on the hyperlink contained within it. Here that conclusion depends upon the search results including, along with the hyperlink and the shortened URL of the Underworld article, the following phrases: "Underworld loses valued friend at court"; "SpecialsGanglandKillings"; and "Pub bouncer‑turned‑criminal lawyer George Defteros always prided himself on being able to avoid a king hit". This contention was accepted by the Court of Appeal.

It may be accepted that, if a search result, by its content, was likely to entice a searcher to choose a given third‑party webpage over others, then an inference might be drawn that the assistance in the search was given with a common intention shared with the third party in the sense of a concerted action to a common end. However, here, even that evidentiary foundation was absent.

Based on the foregoing finding, the words that accompanied the search result simply described a story about a criminal lawyer and his clients, and no more. They indicated what the Underworld article would contain. Of course, for the reasons already given, it may be accepted that the very purpose of the search engine is that it will produce search results that might be accessed by a given user. Depending on what a user is searching for, and why, words that indicate the contents of a given webpage listed in the search results may encourage that person to click on the hyperlink. But this is very far from a basis for a finding of enticement.

The findings in this case about the Google searches of the respondent's name bear this out. The primary judge made detailed findings about seven individuals. Three of these individuals worked at Defteros Lawyers (one of whom was the respondent's wife) and each of them had previously seen the article. The findings do not disclose why they were making these searches. Another individual was the respondent's son; he was prompted by his friends to find the article on the internet. A further individual was a friend of the respondent; he undertook a search to find the respondent's office telephone number. He clicked on the hyperlink to the Underworld article, which he had previously read. Another person had used the respondent in the past as his solicitor. In 2016, he and the respondent discussed "defamatory information on the internet" and the respondent then showed him the search result and the Underworld article. This person later undertook a Google search of the respondent's name, then found the hyperlink to the Underworld article and read it. Why he did this was not the subject of any finding. Finally, another acquaintance of the respondent, whilst searching for the respondent's office telephone number, found the hyperlink to the Underworld article (which he had previously read), clicked on this hyperlink and then read the article (again). Again, no finding was made about why he did this. There was little evidence about the motivations of the other people who clicked through to the Underworld article. The primary judge made a number of unchallenged inferences. Her Honour inferred that a significant number sought out the article in the context of the respondent's complaint to the appellant; some were looking for a lawyer; some were existing clients looking for the contact details of Defteros Lawyers; and a few were considering employment with the firm. Others made searches out of idle interest and curiosity.

Not only is there no basis for a conclusion that the snippet was likely to entice a user to click on the Underworld article, but even if the defamatory imputations in the Underworld article might be said to be enticing, the primary judge made the following finding of fact, which remained unchallenged on appeal: "There was nothing in the search results themselves that incorporated or drew attention to the defamatory imputations that Mr Defteros alleged were conveyed by the Underworld article". (emphasis added)

For these reasons, the facts fall well short of supporting a conclusion that the specific words accompanying the hyperlink to the Underworld article were likely to entice individuals to click on it. The second basis for publication is rejected. The third basis: incorporation

The third basis for publication relied upon the concept of incorporation. In Duffy, where the snippet contained defamatory statements, Kourakis CJ was of the view that this constituted an "incorporation" of that material. His Honour said: "I prefer to speak of incorporation, rather than adoption or endorsement, because lending or imparting weight to the truth of a defamatory imputation is not relevant of the law of defamation in any other context. Incorporation focuses the inquiry on whether the defamatory material is, as a factual matter, incorporated into the publication of the reference or hyperlinker. When referring to another source, the greater the information which is provided about the content of the reference material, irrespective of whether the reference repeats a defamation, the more closely connected the act of reference is to the publication of the referenced material. Indexing by reference to the title and author of the material will only rarely convey sufficient information about the contents so as to constitute a publication of the underlying webpage. However the addition of a snippet, or an abstract, of the material may do so. That is because the searcher only has to assess the snippet or abstracts presented to him or her instead of undertaking the laborious task of going to each reference and assessing them one at a time. A reference accompanied by a snippet or abstract of the defamatory material is even more likely to amount to an incorporation of the hyperlinked webpage. That is because the hyperlink, if used, will direct the searcher to that very material."

It is plain from the foregoing passage that the concept of incorporation that Kourakis CJ had in mind is not limited to the reproduction of defamatory material in a snippet. It would extend to any snippet – that is, words accompanying a hyperlink – that conveys "sufficient information about the contents so as to constitute a publication of the underlying webpage". It was on this basis that the Court of Appeal decided that the Underworld article was "incorporated" by the search result because of the words that accompanied the hyperlink. These words were said to have a "close connection to the parts of the Underworld article dealing with" the respondent and "both directed and encouraged" the reader to click on the hyperlink for further information. By these means, the Google search engine lent "assistance to the publication". A distinction in that respect was, in effect, drawn between the bare expression of a URL in a hyperlink – said to be the circumstance in Crookes – and the position here.

With respect, this reasoning confuses three concepts. To the extent that it suggests that the appellant was a publisher by assisting The Age, this is insufficient to establish publication because, for the reasons given above, the appellant's assistance was not given pursuant to a common intention shared with The Age to publish the article. Moreover, for the reasons set out above concerning the issue of enticement, there is no justification for the proposition that the words accompanying the hyperlink either "directed" or "encouraged" the reader to click on the hyperlink.

To the extent that the reasoning suggests that the appellant was a publisher by incorporation of the content of the Underworld article, the conclusion has no factual basis. The appellant's search engine generates a selection of words or phrases or both from a given webpage, which give the searcher some idea of the content of the webpage. The publication of those words or phrases cannot sensibly be equated with publication of the entire contents – that is, every single word, picture or symbol – of the webpage from which those words or phrases have been taken.

To the extent that the reasoning suggests that the appellant was a publisher by ratification or adoption of the Underworld article, there is also no basis for this conclusion. The inclusion of words or phrases accompanying the hyperlink does not, with great respect, evidence or demonstrate an adoption of, or an assumption of responsibility for, the contents of a given webpage – unless some language of adoption or words that show the taking of responsibility are displayed in the search result. No such language or words may be found in the search result for the Underworld article.

Nor was the provision of notice in the form of the inaccurate removal request – which was relied upon by the respondent as an indicium of publication – sufficient to make the appellant a publisher. To so hold would be to introduce concepts of reasonableness and negligence that are foreign to the law of defamation. As senior counsel for the appellant observed on this appeal, although there was some conflation between publication and the defence of innocent dissemination in the courts below, the false notice given by the respondent ultimately was relevant only in relation to the defence of innocent dissemination. The third basis for publication is rejected. The defences

For the foregoing reasons, the appellant did not publish the Underworld article. It is therefore unnecessary to consider the defences of innocent dissemination and qualified privilege. Conclusion

The appeal should be allowed. We agree with the orders proposed by Kiefel CJ and Gleeson J.