Reading time: 4 min
Key Takeaways
- Direct liability: A German court ruled Google’s AI Overviews are its own statements, not just links, making it directly responsible for false content.
- No escape via user checking: The court rejected Google’s defense that users can verify sources, citing studies showing almost no one clicks through.
- Broader industry impact: This reasoning could extend beyond Google to any AI provider that rewrites web content—ChatGPT, Claude, Perplexity included.
Two Publishers Falsely Linked to Scams
Here’s what actually happened. In June 2026, the Regional Court of Munich ruled on a case that could reshape how we think about AI-generated answers. Two Munich-based publishing companies were falsely associated with subscription traps and shady business practices by Google’s AI Overviews. The AI had mixed data from other companies—real bad actors—with the publishers’ information, creating connections that didn’t exist in any cited source. Let me show you the data: the publishers sent a formal cease-and-desist. Google didn’t respond adequately. Case number 26 O 869/26 landed in court.
The judges made a critical distinction. Classic search engines point to external sites. They make third-party content findable but don’t rewrite it. AI Overviews do something different: they reformulate results “in their own words and according to their own structure,” as the ruling states. In this case, the AI Overview opened with statements like “yes, this company is known for its dubious practices,” then built its own structure—summary, warning signs, user advice—none of which appeared in the underlying search results or linked sources. For the court, these were Google’s “own statements.”
Why Existing Precedent Doesn’t Protect Google
Google leaned on a well-established German Federal Court of Justice (BGH) precedent. That decision holds that search engines and autocomplete are only indirectly liable—they show content produced by others. Imposing systematic verification would break search. The Munich court said this logic doesn’t apply to AI Overviews. Unlike traditional search, the AI generates “independent, new, and substantial statements” by combining and evaluating several third-party sites. Only Google can verify these statements, if only by comparing sources to its output. And the judges noted the AI Overview is “absolutely not essential” for using the internet. Classic results already let users sort information.
Nobody talks about this part: the court also examined the argument that users can check sources themselves. Google claimed the public knows AI content isn’t always accurate. The judges rejected this flat out. The AI Overview was “self-contained” and “an independent statement” whose content could be understood without cross-referencing. Studies show users rarely click through AI Overview sources anyway. The court drew a parallel to press law, where editors are liable for article summaries even if nobody reads the full piece.
Free Speech Protection Reduced for AI Content
This isn’t a take—it’s a pattern. The court also addressed free expression for AI outputs. An AI-generated opinion, the judges said, is “not the expression of a conviction acquired by the speaker, but the result of an algorithm.” Offering an AI-assisted search tool is “primarily an expression of Google’s commercial activity” and only secondarily about expressing opinions or convictions. In the balance between the plaintiffs’ rights and Google’s interests, the company had to yield—especially since the contested claims were based on false facts. The AI had linked the publishers to companies that, under sworn statements, had absolutely no connection to them.
Google lost on nearly all counts. The court ordered it to stop disseminating claims about scams, dubious company ties, subscription traps, fake phone calls, or unavailability. Only two minor requests were denied. The risk of repetition persisted even though the specific false texts weren’t live anymore—Google hadn’t made an irrevocable cessation statement with penalties, and the algorithms could generate the same claims again. Google must pay 80% of legal costs, with the plaintiffs splitting the remaining 20%. The ruling’s impact may extend beyond Germany.
The Technical Detail That Changes Everything
Slow down. Think. An analysis by AI startup Oumi for the New York Times showed Google’s AI Overviews, running Gemini 3, answered correctly 91% of the time. Decent for daily use. But at Google’s scale, that’s millions of wrong answers every hour. Worse: 56% of correct Gemini 3 answers couldn’t be confirmed by the sources Google itself cited. The AI produces answers whose origin users can’t trace.
That’s exactly what the Munich court decided. AI formulates its own statements, which don’t appear in any cited source, and the operator is answerable in court. Will this hold on appeal? Google confirmed it will fight, calling the case about “specific, limited errors, not about the fundamental way AI Overviews display web content.” I’ve seen this play out before. If the reasoning sticks, it could reach ChatGPT, Claude, Perplexity, and any system that rewrites web content. The playbook changed. Again.
Building websites since before Google existed. I’ve run SEO, growth, and content for startups across California — and I’ve watched every ‘revolutionary’ tactic eventually expire. What doesn’t expire: understanding systems, compounding effort, and thinking slower than everyone else.