Munich Court Rules Google Liable for AI Overviews

On May 28, 2026, the Regional Court of Munich (Landgericht Muenchen I) issued a temporary injunction in case no. 26 O 869/26 barring Google from repeating false statements about two Munich-based publishers in its AI-generated "AI Overviews" feature. The Next Web and The Decoder describe it as the first known court decision treating AI Overviews as the platform's own content rather than aggregated third-party results.

Google's AI Overviews had falsely linked the two publishing companies to scams, subscription traps, and "dubious business practices" for certain search queries. According to The Decoder, which reviewed the ruling, the AI mixed up information about genuinely sketchy companies with the plaintiffs and drew connections that appeared in none of the linked sources. One example from the ruling: the AI opened with a confident statement such as "Yes, [company] is known for dubious business practices," then generated its own structure of "red flags" and user tips. The plaintiffs sent Google a cease-and-desist letter that Google did not respond to appropriately, per The Decoder.

The court found that AI Overviews work nothing like traditional search results. Unlike conventional results that index and link to external pages, AI Overviews "rewrite and judge results in their own words and according to their own structure," per The Decoder's translation of the ruling. The court identified this as producing "independent, new, and substantive statements" - and Google is the only party that can verify those statements "at least by comparing the underlying third-party websites with its own statements based on them."

The court also ruled that Germany's Federal Court of Justice (BGH) precedents giving traditional search engines limited "indirect infringer" liability do not apply to AI Overviews, because those precedents depended on search engines merely making third-party content findable - not generating new claims. Google built the AI and "alone has influence over the AI's offering and the algorithms with which the AI operates," the court stated, per The Decoder.

At the hearing, Google argued that users could verify AI summaries by checking linked sources and that users generally knew "that information generated with AI should not be blindly trusted," per The Decoder. The court rejected this. It noted that the AI overview was "understandable on its own" with "a self-contained statement" and "no reference to other possible interpretations or even unreliable content." A separate Pew study cited by The Decoder found that only about 1 percent of users click a source link directly from AI Overviews, supporting the court's view that most users accept the summary as authoritative.

The court drew a parallel to press law: publishers can be liable for teasers understandable on their own even if readers never read the full article. The court also noted a "protection gap" - the third-party websites whose content served as sources had not made the false statements, so victims had no effective recourse against them; under prior rules they also could not effectively sue Google. Google therefore could not invoke host-provider protections under the Digital Services Act.

Google also did not issue a cease-and-desist declaration with a penalty clause, leaving the risk of repeated violations open. The court ordered Google to cover 80 percent of legal costs; the two plaintiffs pay 10 percent each.

An analysis by AI startup Oumi cited by The Decoder found that Google's Gemini 3 model answers correctly in approximately 91 percent of AI Overview queries. At Google's scale, a 9 percent error rate still means millions of incorrect answers served daily. The same analysis found that 56 percent of correct Gemini 3 answers could not be independently traced to the sources Google linked - the AI states things no linked source actually says. The court's ruling addresses exactly this gap.

The injunction is preliminary; Google has not publicly commented and can appeal. The court noted the ruling may have international reach. Industry observers will track whether appellate courts adopt the same reasoning; whether EU-level bodies offer guidance under the Digital Services Act; follow-on litigation using similar theories in other EU countries; and whether answer-engine providers running ChatGPT, Claude, or Perplexity adjust provenance presentation, language hedging, or remediation workflows in response.