A German court just did what no regulator has managed to pull off: it held Google directly responsible for what its AI says about people and companies.
The Regional Court of Munich issued a preliminary injunction against Google on June 9, 2026, ordering the company to stop spreading false claims through its AI Overviews feature. Two Munich-based publishing companies found themselves falsely linked to scams, subscription traps, and "dubious business practices" by Google's AI. The AI had mixed up information about genuinely shady companies with the plaintiffs and drew connections that appeared nowhere in the linked sources.
Google argued that users know AI makes mistakes and should verify the results themselves. The court wasn't buying it.
The Argument That Fell Apart
Google's defense was simple: we're just a search engine. The AI Overview summarizes what's already out there. If it gets something wrong, users can click through and check the sources.
The court dismantled this piece by piece.
First, the AI Overview didn't just summarize. It generated "independent, new, and substantive statements" by evaluating and combining content from multiple sources. In this case, the AI opened with confident claims like "Yes, [company] is known for dubious business practices" and then built its own structure with a summary, red flags, and tips for users. None of the linked pages made those connections. The AI invented them.
Second, the court pointed out that traditional search engine liability protections exist because search engines merely make third-party content findable. They're a pointer, not a publisher. AI Overviews work nothing like that. They rewrite, judge, and create new content "in their own words and according to their own structure."
Third, and most devastating: the court said AI-generated content gets less free speech protection than human expression. An AI's output is "not the expression of an acquired conviction of the persons expressing it, but the result of an algorithm." It's a commercial product, not an opinion.
Where Search Immunity Ends
For decades, search engines have operated under a simple legal framework: indexing the web at scale means some harmful content slips through, and that's unavoidable. The search engine isn't endorsing anything. It's just showing you what's out there.
The Munich court said that logic stops at the AI Overview box.
When Google takes information from multiple sources, rephrases it, adds its own structure, and presents it as a confident answer at the top of the page, Google becomes the content provider. The distinction matters because content providers are liable for what they publish. Search engines aren't.
The court also addressed a protection gap that would exist if AI Overviews received search engine immunity. The third-party websites whose content Google's AI drew from hadn't made the defamatory statements. So the victims couldn't sue the sources. And under existing rules, they couldn't effectively sue Google either. Someone had to be responsible. The court decided it would be Google.
Google picked up 80 percent of the legal costs. The publishers won on almost every count.
The Numbers Behind the Nightmare
The scary part isn't this one case. It's what happens when you multiply the error rate by Google's scale.
An analysis by AI startup Oumi, reported by the New York Times, found that Google's AI Overviews are accurate about 90 percent of the time. That sounds fine until you remember that Google processes more than five trillion searches per year. A 10 percent error rate means tens of millions of wrong answers every hour.
And the errors aren't random noise. Pew Research found that when an AI Overview appears on a search results page, clicks to traditional results drop from roughly 15 percent to 8 percent. People see the AI answer and stop looking. They don't click through to verify.
Even when the AI gets the answer right, it's often not backed by the linked sources. The Oumi analysis found that 56 percent of correct answers were "ungrounded", meaning the websites Google cited didn't fully support the information the AI provided. The AI is giving confident answers whose origins users can't trace, and most users never try.
What This Means for Brands
This ruling is preliminary, and Google will almost certainly appeal. But the signal is clear, and it's going to travel.
If the Munich court's logic holds, every company that builds generative AI search tools faces legal exposure when those tools produce false or defamatory claims. That includes Google, Microsoft, OpenAI, Perplexity, and anyone else serving AI-generated summaries at scale.
For brands, this changes the calculus around AI search visibility. Until now, the conversation has been about whether your brand shows up in AI answers at all. The new question is whether your brand shows up accurately, and what you can do when it doesn't.
The publishers in this case sent Google a cease-and-desist letter. Google didn't respond appropriately, and that inaction became part of the ruling. The court noted that since only Google can adjust the models and algorithms that create AI Overviews, only Google can reliably stop the system from repeating the same falsehoods.
That puts the burden on the platform. But it also means brands need to monitor what AI search tools are saying about them and document everything. Screenshots, search terms, timestamps, and records of every communication with the platform.
The Precedent That Travels
The Munich ruling has international reach, according to the court. Even though it's a German decision, the reasoning applies anywhere AI search tools operate.
The court said something that should make every AI search provider uncomfortable: AI summaries are "by no means absolutely necessary" for using the internet. Traditional search results already help users sort through information. The AI Overview is an extra feature, and extra features don't get the same legal immunities as core infrastructure.
That distinction matters because it reframes the entire liability debate. We're not talking about whether AI search should exist. We're talking about whether companies that choose to deploy it get to offload the risk onto the people and businesses their AI misrepresents.
The answer, at least in Munich, is no.
If other courts follow this logic, and they will, the companies deploying generative AI search will have to build real accountability into their systems. Not disclaimers. Not "users should verify." Actual accountability for what their products say about real people and real businesses.
That's not a bug in the system. It's the system finally catching up.