Google Found Liable for AI Overviews' False Claims in Germany

A German regional court has drawn a line that AI search products have been hoping to avoid. On June 10, Munich's Regional Court issued a temporary injunction finding Google directly liable for false statements in its AI Overviews, classifying the summaries as the company's own editorial output rather than neutral links to external sources, The Decoder reports.

The legal reasoning is the part worth paying close attention to. The court characterized AI Overviews as generating "independent, new, and substantive statements" and concluded that Google "alone has influence over the AI's offering and the algorithms with which the AI operates." That distinction strips the liability shield traditional search engines relied on. Google argued that users could verify claims themselves through linked sources, but the court rejected that logic: the possibility of disproving a statement through further research does not regularly exempt a party from liability for making it. The two Munich publishers at the center of case no. 26 O 869/26 had been incorrectly linked by Google's AI to scams, subscription traps, and fraudulent business practices, with the court finding the AI had mixed up information from other companies and applied it to the plaintiffs.

The numbers cited in the ruling add weight to that conclusion. Google's AI Overviews answered correctly around 91% of the time in one analysis, which sounds reassuring until you account for scale. That same analysis found 56% of the correct Gemini 3 answers couldn't be backed up by the sources Google linked. And according to The Next Web, barely 1% of users actually click the source links in AI Overviews, which the court used to undercut the "just verify it yourself" defense. Google was ordered to cover 80% of legal costs.

The honest caveat is that this is a temporary injunction. Google confirmed it will appeal, and a higher court could draw the line between AI publishing and search hosting differently. What the reporting also does not fully resolve is whether this creates liability for each incorrect AI answer individually or only for identifiable false statements about specific named parties, a scope question that will shape how companies actually respond.

For businesses whose reputations have been damaged by AI mischaracterization, this ruling offers a working legal template, at least in Germany. Companies that built AI answer products with tight source attribution and prominent citation display, rather than fluent summaries disconnected from provenance, are better positioned if this standard spreads. For everyone else building on AI-generated answers, the message from Munich is plain: "the AI said it" is not a defense.