1. A changed search experience
Until less than two years ago, searching online still followed a familiar ritual: type a query into a search bar, scan a list of «blue links», click through to the source. Google Search has long played a central role in this ritual, powered by a system able to retrieve up-to-date content from a vast pool of data, respond almost instantly, and deliver the same experience to billions of users across the globe.
For a long time, Google had no reason to change course. Its revenue depended on the sheer volume of searches performed every day: each query generated a chance to place an ad, and each interaction generated data that Google could analyse to refine targeting and keep advertisers engaged. A conversational, AI-driven search experience, like AI Overviews, threatened this model at its core: if users could get a direct answer in one step, they would search less, leaving fewer opportunities for ad placement and less data to work with. Disrupting a system that worked this well made little economic sense.
There was also a reputational risk to consider. Google’s research labs had developed advanced conversational language models well before OpenAI’s public launch of ChatGPT, suggesting that technology itself was not an obstacle. On the other hand, deploying a system prone to hallucinations would have undermined the credibility that Google had built over decades and depended on.
The turning point came in late 2022. Faced with the rapid rise of generative search, Google set aside its cautious approach to defend its position in the market. AI Overviews were the result: instead of a list of links, users would now find an AI-generated synthesis placed at the top of search results, merging fragments from multiple sources into a single, authoritative-looking answer.
This is not a simple product update, but a fundamental redefinition of Google’s legal and economic role. A traditional search engine only indexes content created by others, leaving it to the user to reach the source. AI Overviews do the opposite: rather than pointing users elsewhere, the system pulls information from a range of sources, reworks it into a single narrative, and delivers it as if it were speaking on its own authority. This is the point at which it becomes legitimate to ask whether Google has shifted from being a neutral messenger to being, in effect, an author.
2. The Munich Ruling: Google held liable for its AI Overviews
The most significant legal challenge to this new model arrived on 28 May 2026, when the Regional Court of Munich I (Landgericht München I) issued its judgment in case 26 O 869/26. For the first time, a German court held that the provider of an AI system could be directly liable for the statements generated by the system, rather than treating it as a mere intermediary passing along third-party content.
The case involved two Munich-based publishing houses that discovered AI Overviews had linked their names to fraudulent schemes and disreputable business practices. When users searched for the publishers’ names alongside terms like «fraud scheme», the system produced a confident, negative summary, complete with self-generated headings on warning signs and red flags. The core accusation — a link between the plaintiffs and an unrelated company under scrutiny for questionable practices — traced back to no identifiable source: even the article AI Overviews cited to support it made no mention of the plaintiffs at all. In effect, the system had fabricated a connection that existed in no single source, and presented it as established fact.
Google’s defence rested on a well-established line of reasoning: the underlying data came from third-party websites, the system merely summarised it automatically and users remained free to check the sources themselves. The court dismissed this reasoning on three grounds. First, it found that AI Overviews produce independent, substantive content attributable to Google itself, not simply a reproduction of third-party material, placing the company, in this respect, closer to a publisher than to a neutral intermediary. Second, the court held that established case law from the German Federal Court of Justice (Bundesgerichtshof), which had historically limited the liability of search engines, does not extend to AI Overviews. The distinction lies in what each system does: a traditional search engine surfaces content without adding to it, while AI Overviews build new statements by selecting, evaluating, and restructuring information according to criteria set by Google itself. Third, the fact that a false claim could, in principle, be uncovered through further research does not shield the party that published it from liability. The court also flagged a broader problem: the third-party sites behind the summary had never made the disputed claims themselves, so they could not be sued for them. Limiting Google’s liability to obvious infringements only would leave a gap in legal protection, since those harmed by the AI’s own statements would be left with no recourse at all.
Combined, these findings led the court to exclude the possibility that Google could rely on the protections the Digital Services Act (DSA) reserves for hosting providers — protections that apply only to platforms hosting content created by others, not to those generating it. The court’s injunction prohibited Google from repeating the specific false claims, with fines of up to €250,000 per violation. It also noted that the ruling would apply beyond German borders, under Article 36(1) of the Brussels I Regulation.
Overall, the reasoning opens a significant gap in the liability framework: anyone integrating an LLM into a product aimed at the European public may now be treated as a publisher of the content it generates, not as a search engine. The limited-liability regime that shielded Google, Bing, and other aggregators until now does not automatically extend to content the system itself produces.
The DSA framework has also served as the basis for AGCOM’s assessment of Google’s generative AI services. In particular, on 29 April 2026, the Italian Communications Regulatory Authority acting as the country’s Digital Services Coordinator, referred Google’s AI Overviews and AI Mode to the European Commission under Article 65 of the DSA, following a complaint by the Italian publishers’ federation FIEG. AGCOM raised concerns echoing those at stake in Munich, including the reduced visibility of editorial content, the risks faced by smaller and independent publishers, and the provision of unverifiable or fabricated answers, identifying them as potential breaches of Articles 27, 34 and 35 of the DSA, which govern recommender-system transparency and the mitigation of systemic risks (AGCOM, 2026).
3. The impact of AI-generated search on publishers and emerging legal challenges
The Munich ruling did not emerge in a vacuum. It intersects with a broader structural problem that competition law is only beginning to address: the zero-click effect on digital publishers.
Historically, search engines and publishers existed in a symbiotic relationship: publishers provided content, and Google, like other search engines, provided traffic. AI Overviews break this cycle, creating a scenario where website operators are effectively «forced to compete against themselves» (Lehmkühler and Steinseifer, 2026). The mechanism is paradoxical: for a website to appear in regular Google search results, it currently has no choice but to allow Google to use its content for AI summaries, which in turn reduces the reason for anyone to visit the site in the first place. Publishers are left in a loop in which their own content is turned against the sustainability of their business.
It should be noted that AI Overviews do not appear for every query. Google states that they are triggered only when its system judges a generative summary «especially helpful» against the traditional list of links, an automated assessment that varies by query type and carries no additional SEO requirement. Nevertheless, such selective, unpredictable triggering leaves open the legal questions it raises, making scrutiny of the practice legitimate.
The European Publishers Council has filed a formal complaint with the European Commission, arguing that Google is abusing its dominant position in general search through the deployment of AI Overviews and AI Mode within Google Search, in breach of Article 102 TFEU (European Publishers Council, 2026). Scholars have traced this possible violation back to how search visibility is being used as leverage: appearing in the regular results is effectively made conditional on allowing that same content to be used, without a genuine opt-out or any compensation, to generate AI Overviews.
Along similar lines, French Competition Authority had already sanctioned Google over the undisclosed use of press content to train its AI systems (Autorité de la Concurrence, 2024), and the European Commission has opened its own investigation into how publisher and video content feeds into Google’s AI features (European Commission, 2025) — an inquiry the EPC complaint now reinforces.
What is taking shape is a shift playing out across different legal fields at once. It is not only a question of whether Google’s status under the DSA should change, as the Munich ruling suggests. The same conduct is also being tested against competition law, specifically through the exploitative and self-preferencing arguments emerging around the practice, and against copyright and related rights, through the unauthorised use of publisher content for training. In every case, the issue is the same: the legal categories built to protect a passive intermediary no longer hold for how Google’s AI-powered search now operates.
4. The UK’s answer: a genuine opt-out for publishers
On 3 June 2026, the UK’s Competition and Markets Authority (CMA) imposed a new conduct requirement on Google’s search business under the UK’s digital markets competition regime, following Google’s designation with strategic market status in general search. In what the CMA itself calls a world first, publishers will be able to opt out of having their content used to power AI features such as AI Overviews, without losing their ranking in standard search results. Google must also let them opt out of having their content used to fine-tune its AI models, ensure clear attribution of publisher content in AI-generated results, and publish regular compliance reports with detailed engagement metrics (Competition and Markets Authority, 2026).
The measure responds directly to the problem discussed above, where opting out had meant disappearing from search altogether: by separating participation in AI features from participation in ordinary search results, the CMA gives publishers a genuine choice rather than a binary one between full exposure and full invisibility. This, in turn, strengthens their negotiating position over how their content is extracted, synthesised, and displayed. As the first concrete step of its kind, the conduct requirement may shape how other jurisdictions approach the same problem.
5. Conclusion: towards a new legal framework for a new era of search
Google has already made clear it will contest the Munich decision, calling the case narrow in scope and stating it «disagree[s]with the ruling and plan[s]to appeal». Since this is only a preliminary injunction, the matter is far from settled. But the logic behind it isn’t specific to Google: if courts in other jurisdictions embrace the same approach, any provider whose AI systems generate summaries of online content could face the same exposure.
The case exposes the limits of the existing regulatory architecture. Instruments such as the Digital Services Act govern intermediary services, including very large online search engines, on the premise that they host, index, and facilitate access to content created by others. That premise no longer fully reflects the operation of AI-powered search, which does not merely retrieve information but produces it, presenting its outputs with the authority of a brand users already trust.
This shift calls for a legal framework capable of addressing this new way of searching for information online. Moreover, the issues raised by AI Overviews extend beyond platform liability: they intersect with copyright, competition law and other areas of digital regulation, revealing a set of significant interconnected questions. What emerges, in other words, is not a gap to be patched but a broader regulatory rethinking to be undertaken — one capable of keeping pace with the distinctive nature of generative search.
