German Court Treats Google AI Overviews as Google’s Own Words

A regional court in Munich has drawn a sharper line around AI search liability. In a temporary injunction, the court treated Google’s AI Overviews as Google’s own generated statements, not as ordinary search results or neutral links to third-party pages. The case involved two Munich publishers that were wrongly associated with scams, subscription traps, and dubious business practices in AI-generated search summaries.

The important detail is not simply that an AI answer was wrong. According to the reporting, the contested overview created connections that did not appear in the linked sources. That made Google’s “users can check the links themselves” defense much weaker. The court reasoned that the overview was understandable on its own, had its own structure, and could not be treated like a traditional list of search results.

That is why the Hacker News thread became less about benchmark accuracy and more about deployment responsibility. At search scale, even a small error rate can produce many damaging claims. If a model synthesizes a confident accusation from source fragments, the harmed party may not have a meaningful claim against any linked publisher because the linked publishers did not make the statement. The platform that generated and displayed the answer becomes the obvious legal target.

The original report came from The Decoder, while the HN discussion drew hundreds of comments within the 72-hour window. The ruling may still face appeal and jurisdictional limits, but it gives AI search operators a concrete warning: summaries that read like direct answers may also carry direct responsibility.