German court holds Google liable for defamatory AI Overviews as its own speech

The Regional Court of Munich has issued a temporary injunction against Google after its AI Overviews falsely associated two Munich publishers with scams, subscription traps, and fraudulent business practices — connections that appeared in none of the cited sources. The court’s key move was classifying the AI-generated summaries as Google’s own content rather than third-party material, making the company a direct infringer. Established German case law that shields search engines and autocomplete from liability doesn’t apply, the court reasoned, because a conventional search engine merely points to external sites, while AI Overviews synthesize new, standalone claims that only Google is in a position to verify against the underlying sources.

Google’s defenses fared poorly. Its argument that users should fact-check AI output themselves was rejected: the overviews read as self-contained, authoritative statements, and research shows users rarely click through to sources. The court drew an analogy to press law, where publishers answer for teasers regardless of whether anyone reads the full article. Google also couldn’t claim host-provider protections under the Digital Services Act or rely on standard notice-and-takedown procedures. Notably, the court held that AI-generated statements deserve weaker free-speech protection than human expression, since they reflect an algorithm’s output and Google’s commercial interests rather than anyone’s genuine convictions. Google was ordered to cover 80 percent of legal costs, and the injunction stands because nothing prevents the system from regenerating the same false claims.

The ruling’s significance extends well beyond one case. Independent analysis cited in the article found Gemini 3-powered Overviews answer correctly about 91 percent of the time — which at Google’s scale still translates to millions of errors hourly — and that over half of even correct answers can’t be traced to the linked sources. If the Munich court’s logic survives appeal and spreads to other jurisdictions, every operator of AI summarization systems, including ChatGPT, Claude, and Perplexity, could face direct liability for hallucinated claims about real people and businesses.