On May 28, 2026, the Munich Regional Court issued a temporary injunction barring Google from spreading false information about two Munich-based publishers via AI Overviews in search results (case no. 26 O 869/26). The ruling was announced on June 9, 2026 — and may reshape how AI platforms are held liable for generated content.
Key takeaways
- Munich Regional Court classified AI Overviews as Google's own content, not indexed results
- Google held directly liable for falsely linking two publishers to scams and dubious practices
- Existing search engine liability case law rejected — AI Overviews generate independent claims
- Google's argument that users can verify sources themselves dismissed — only 1% of users click sources in AI Overviews
- Ruling may extend to ChatGPT, Claude, Perplexity, and other systems summarizing web content
What AI Overviews got wrong
Google's AI Overviews incorrectly linked two Munich-based publishing companies to subscription traps, dubious business practices, and clone companies. The court found that the AI mixed up information about genuinely sketchy companies with the plaintiff publishers — and built connections that did not appear in any of the linked sources. AI Overviews opened results with phrases like "Yes, [company] is known for dubious business practices," then generated its own structure with a summary, red flags, and user tips. The publishers sent Google a cease-and-desist letter — the company did not respond appropriately.
Why this is not an ordinary search engine
The core legal distinction concerns the liability model. German case law from the Federal Court of Justice (BGH) granted traditional search engines and autocomplete limited liability — a search engine merely points to external content, and imposing a proactive duty to verify results would threaten how search engines function.
The Munich court ruled that this reasoning does not apply to AI Overviews. A conventional search engine points to external websites. AI Overviews generates "independent, new, and substantive statements" by evaluating and combining content from multiple external sources. The court found that the AI rewrites results "in its own words and according to its own structure" — and only Google can verify those claims, by comparing the linked websites with its own statements derived from them.
The ruling states directly: "only the defendant has influence over the AI's offering and the algorithms with which the AI operates." This means Google builds, offers, and is responsible for what the technology produces.
Hosting provider protections and DSA do not apply
The court also examined whether Google could invoke host provider protections under the Digital Services Act (DSA) — and rejected this. AI Overviews is not a service storing third-party content. It generates its own statements synthesizing information from external sources, which excludes application of the standard notice-and-takedown procedure applicable to search engines.
The court identified a protection gap: companies whose reputation was harmed could not sue the sources (which did not contain the disputed claims) nor effectively sue Google under existing liability rules. The ruling closes that gap.
The scale problem: 9% error rate at Google's traffic volume
An analysis by AI startup Oumi commissioned by the New York Times found that Google's AI Overviews with the current Gemini 3 model answers correctly 91% of the time. At Google's scale, even 91% accuracy leaves several percent wrong answers served to tens of millions of users per day — millions of potentially false claims per hour.
More importantly, the analysis found that 56% of Gemini 3's correct answers could not be corroborated by the sources Google linked. The AI gives answers whose evidentiary trail users cannot trace. Pew Research data shows that only 1% of users click on sources in AI Overviews.
Why this matters
The Munich ruling is the first judgment to explicitly break the legal boundary between a search engine and a generative AI system. It is not a BGH appellate ruling — it may be appealed, and Google has not commented publicly. But if this logic is upheld or adopted by other jurisdictions, the consequences extend well beyond one dispute.
Google, ChatGPT, Claude, Perplexity, Microsoft Copilot — all these systems deliver web content summaries at scale. Under the legal model proposed by the Munich court, each of them is the author of its claims, not a distributor of others' content. That changes the risk calculation for AI providers entering markets with strong personal reputation protection frameworks.
The ruling also contains a significant remark on freedom of expression: an AI's opinion "is not the expression of an acquired conviction of the persons expressing it, but the result of an algorithm." The court classified this as Google's commercial activity, not protected speech.
What's next?
- Google has the right to appeal — the appellate outcome will determine whether the Munich reasoning becomes binding
- The European Commission and national DSA enforcement bodies may cite this ruling in proceedings under the DSA and GDPR
- Other AI companies offering generative search (Perplexity, Bing Copilot, SearchGPT) should assess their legal exposure under civil law systems in light of this ruling
