From Search Engine to Answer Engine: Why Google’s AI Future Threatens the Legal Foundations of the Open Web

Introduction : The deal was easy, but no one signed anything. The users clicked through, Google searched it, and publishers wrote it. Traffic went to sources; advertising money went to the eyeballs; and Google’s incredible market force was built on the premise that it was a map of the web, not a place to go. This flawed, disputed, but internally consistent agreement prevailed for about 20 years. Now, AI Overviews has come into the picture, and that deal is starting to fall apart. Users who get the answers they’re looking for right on Google’s search results without ever clicking on a third-party site have become the rule rather than the exception, and that’s called Zero-Click Searches.

Now, Gemini can craft a sophisticated, confident response to inquiries on a variety of subjects, from drug interactions to mortgage rates to historical events to travel visa requirements or symptoms of the condition you’d rather not mention in public. Google used to send you somewhere; now, more and more, Google tells you. But this change is not simply a change of product. Instead, this is a change in legislation. By moving from an indexing model to a content creation model, Google has abandoned its neutral information intermediary role and consciously made a transition towards that of a publisher.

The Illusion of the “Librarian” Defense

It should be stated that the historical approach to solving the problem of liability of search engines was based on a certain fiction, according to which the information is viewed as something separate from production. This point of view was widely used by the courts of American and European states within the last several decades. Search engines could enjoy such safe havens as platforms because of Section 230 of the Communications Decency Act (CDA) and the European E-Commerce Directive. The search engines could not be blamed for the content they were forwarding, because it was not produced by them.

This distinction was respected by traditional search. Google searched through the semantic layer of billions of websites, sorted them out, and then gave the user a ranked list of doors. The content itself was elsewhere and was created by somebody else, and all that Google had to do was simply stand by the window frame. However, things are very different in the case of Gemini. Gemini does not extract something from the cardiologist’s paper. It deconstructs it, cross-references it with other papers, evaluates their authority, and then constructs it all into its own sentences, its own structure, and its own authoritative discourse. Thus, the result is not a link but a statement.

There we have the Derivative Works doctrine, as quietly as ever, but with the same force as ever. It is not only verbatim copy that is protected under copyright, but it is also substitution for an original work in the market. If the synthesized work generated by the computer answers the same question for which the original article was written, then it is not merely a case of plagiarism, it is a competition as well. Google contends that the search engine produces a transformative work in the sense that it doesn't copy but aggregates, reorganizes, and recontextualizes. That may have been true in some situations, but as soon as the transformation is done well enough so that the user loses motivation to go back to the source, the case fails. Google wants the position of a librarian without any liability.

Market Substitution and the Fair Use Failure

The fourth factor of fair use, which has always been the ultimate battlefield, is the effect of the use upon the potential market for the copyrighted work. And that is where AI overviews are most vulnerable. For decades, courts have recognized that the analysis of fair use should pivot around the concept of Market Substitution, the more secondary uses substitute for the demand for the original, the weaker is the fair use defense.

The substitution that AI Overviews provide is not coincidental. If a publication in health journalism dedicates half a year creating an elaborate guide on how to control Type 2 diabetes with food, it does not just “lose a click,” when Gemini summarizes all of the important parts of that guide into three concise paragraphs. Instead, it loses views, income from advertising, the chance of getting subscribers, and the algorithmic advantage of being easily discovered. Multiply that damage by the number of the categories where AI Overviews are most complete : personal finance, travel planning, consumer electronics, law explanations, and healthcare.

Nevertheless, the problem is not just limited to copyright but has grown to include issues pertaining to competition law as well. If publishers want to avoid having their work included in the AI-powered summaries on Gemini, then they have no option but to forego their participation in the training process of Gemini and therefore lose out on their Google search ranking, or continue participating and allow their intellectual output to be used by their rival company, who provides its products for free. This is exactly the sort of structural inequality that competition law is supposed to investigate. After all, the power that Google possesses is intrinsically tied to its position in the market. Every major publisher cannot afford to give up on Google Search ranking.

The June 2026 Munich Precedent

However theoretical in nature this argument might have been, it became firmly grounded in judicial precedent in June of 2026, with the Munich Regional Court issuing a preliminary injunction that may turn out to be one of the most important cases in the history of platform law. The case emerged after two German publishers sued Google for creating false statements about their editorial works using the AI Overviews in connection with their involvement in fraudulent subscriptions. Importantly, the disinformation did not stem from a single piece of information, but rather from a confident construction based on individual signals that did not support the conclusions reached.

The defendant’s arguments were reasonable and could hold in an earlier phase of legal practice. Namely, Google argued that each of the AI Overviews was linked to sources, allowing individuals to confirm the truth of each claim made. However, the Munich court did not accept such arguments, and with good reason. The fundamental rule that the court identified, accordingly, was that since AI-generated text organizes and arranges information in its own language, it generates independent, original, and substantive statements that make the generating entity liable as an author thereof. Source linking, the court found, cannot shift that liability to the user just like a newspaper cannot absolve itself from liability for an incorrect headline by including a bibliography.

However, the scope of this logic goes well beyond the defamation provisions in German law. Given that AI-generated text represents independent authorship, which entails creating new statements and not merely reproducing third-party material, the Publisher Liability comes into play in a way that has never applied to ordinary search services. According to the present analysis, Google cannot be viewed as reflecting the information available on the Internet. Rather, it generates independent information derived from, but separate from, its sources. This logic destroys the conduit defense precisely because it eliminates the distinction between conduit and author. Every legal system where such a distinction is recognized therefore has the basis for treating AI Overviews as publication material.

The AI Echo Chamber and the Road Ahead

Nevertheless, there is an additional question that has not yet been fully addressed in the discourse of law. What will be the state of the information ecology if the creation of original material is no longer economically viable? As AI increasingly takes away the web traffic that would otherwise go to original work, it begins to destroy the economic incentives to create it. Future AI systems will have to analyze a web increasingly composed of AI summaries of other AI summaries, thus creating a feedback loop of failure. The AI Echo Chamber effect is not a futuristic metaphor but rather the inevitable conclusion drawn from current incentives.

It means that there will be a choice between two futures. The first one is where Google recognizes the fact that its role is now that of a publisher rather than just an index, and sets up licensing models for its relationships with the creators and the publishing industry as a whole. It will mean a complete rethinking of the economic model of the web in line with what we see in the broadcasting industry in relation to music and radio. Early and flawed attempts have been made in Australia through the News Media Bargaining Code and the European Copyright Directive in terms of the “neighboring rights.”

The second alternative sees Google fighting against the responsibilities of a publisher via the courts, via legislation, and by letting individual copyright cases eat away at intermediary protections. The result of this approach is the building of pressure through regulation on multiple jurisdictions, increasing antitrust concerns about the dominant position of Google’s search engine, and the creation of case law, of which Munich is merely the first, that increasingly wears down intermediary defenses that have protected the company for years now.

What it comes down to, what becomes clear through the Munich decision and through the development of AI Overviews, is that this reclassification is not an issue of choice. It is simply the natural consequence of how technology works. What is most significant about AI Overviews is not their impressiveness, but their authority. They are statements, not pointers. Answers, not directions. The point where Google stopped organizing information and started giving answers is the point where Google stopped being the librarian of the Internet. And it became one of its major publishers.

Author: Maitraiy Soni, in case of any queries please contact/write back to us via email to chhavi@khuranaandkhurana.com or at  Khurana & Khurana, Advocates and IP Attorney.

References

Legislation and Regulatory Instruments

1. 17 U.S.C. § 107 (2018). Fair Use, United States Copyright Act. U.S. Congress.

2. 47 U.S.C. § 230 (1996). Protection for Private Blocking and Screening of Offensive Material, Communications Decency Act. U.S. Congress.

3. Directive 2000/31/EC of the European Parliament and of the Council of 8 June 2000 on Certain Legal Aspects of Information Society Services, in Particular Electronic Commerce, in the Internal Market (E-Commerce Directive). Official Journal of the European Communities.

4. Directive (EU) 2019/790 of the European Parliament and of the Council of 17 April 2019 on Copyright and Related Rights in the Digital Single Market (DSM Copyright Directive). Official Journal of the European Union.

5. Treasury Laws Amendment (News Media and Digital Platforms Mandatory Bargaining Code) Act 2021 (Cth) (Austl.) [Australia’s News Media Bargaining Code].

Case Law and Judicial Decisions

1. Authors Guild v. Google, Inc., 804 F.3d 202 (2d Cir. 2015). United States Court of Appeals for the Second Circuit. (Transformative use; fair use defense for large-scale digitization projects.)

2. Campbell v. Acuff-Rose Music, Inc., 510 U.S. 569 (1994). United States Supreme Court. (Defining transformative use and the four-factor fair use analysis, with particular emphasis on market substitution.)

3. Kelly v. Arriba Soft Corp., 336 F.3d 811 (9th Cir. 2003). United States Court of Appeals for the Ninth Circuit. (Indexing and thumbnail reproduction by search engines; early intermediary status analysis.)

4. Landgericht München I [Munich Regional Court I], Preliminary Injunction, Case No. [unreported at time of publication], June 2026. (AI-generated text as independent authorial statement; Publisher Liability for AI Overviews; rejection of source-link defense.)

5. New York Times Co. v. Microsoft Corp. et al., No. 1:23-cv-11195 (S.D.N.Y., filed Dec. 27, 2023). (Pending litigation; copyright infringement claims regarding large language model training on news content and AI-generated substitution for journalism.)

Academic and Policy Literature

1. Balganesh, S. (2011). ‘The Obligatory Structure of Copyright Law: Unbundling the Wrong of Copying.’ Harvard Law Review, 125(6), 1664–1728.

2. Grimmelmann, J. (2015). ‘The Google Dilemma.’ New York Law School Law Review, 53(4), 939–950.

3. Lemley, M. A., & Reese, R. A. (2004). ‘Reducing Digital Copyright Infringement Without Restricting Innovation.’ Stanford Law Review, 56(6), 1345–1434.

4. Levin, P. (2024). ‘Generative Search and the Erosion of the Traffic Bargain: Rethinking Platform Neutrality in the Age of AI Summaries.’ Journal of Internet Law, 27(9), 1–18.

5. Massaro, T. M., & Norton, H. (2024). ‘Siri-ously? Free Speech Rights and Artificial Intelligence.’ Northwestern University Law Review, 110(5), 1169–1246.

6. Murray, M. D. (2023). ‘Generative and Transformative AI and the Future of Authorship.’ Hastings Communications and Entertainment Law Journal, 45(2), 221–278.

7. Samuelson, P. (2023). ‘Generative AI Meets Copyright.’ Science, 381(6654), 158–161.

8. Wu, T. (2024). The Digital Cartel: Search Dominance, Data Leverage, and the Illusion of Platform Neutrality. Columbia Law School Working Paper, No. 24-17.

Journalism, Industry Reports, and News Sources

1. Autorité de la concurrence [French Competition Authority]. (2022, July 1). Decision No. 22-D-13 Concerning Practices Implemented by Google in the Press Publishers Sector. (Neighboring rights enforcement; compensation obligations for news snippets in search results.)

2. Batelle, J. (2024, October). ‘The Collapse of the Web’s Traffic Economy.’ WIRED. Retrieved from https://www.wired.com.

3. Bond, S. (2024, May 14). ‘Google’s New AI Overviews: What They Mean for Publishers.’ NPR Technology. Retrieved from https://www.npr.org.

4. European Commission. (2024). Digital Markets Act, Designation and Obligations for Google Search as a Gatekeeper. DG Competition, Brussels.

5. Fishman, E. (2024, June). ‘Zero-Click Searches Now Account for 65 Percent of Google Queries.’ SparkToro Research Report.

6. Heilweil, R. (2025, February). ‘When Google Gets It Wrong: The Defamation Risk of AI Search Answers.’ MIT Technology Review. Retrieved from https://www.technologyreview.com

7. Reuters Institute for the Study of Journalism. (2025). Digital News Report 2025. University of Oxford.

8. The Markup. (2025, January 9). ‘How Google’s AI Summaries Are Redirecting Traffic, and Where It Goes.’ The Markup. Retrieved from https://themarkup.org.