A German regional court has held Google responsible for a response from its AI that associated accurate facts with a doctor, but which belonged to another man. In a ruling dated September 24, 2026, the Aachen Regional Court ordered Google to cease disseminating three versions of an AI Mode response concerning this doctor.
This chief physician, a professor of neuroradiology, had left his position at the end of June. When internet users searched for the reasons for this departure, AI Mode provided them with a serious criminal accusation. Follow-up queries, using terms like ‘suspension’, ‘dismissal’ or ‘legal proceedings’, resulted in multi-paragraph narratives about judicial consequences. These proceedings did exist, but they concerned another man, also a former university professor, and the court was able to confirm the confusion from its own archives.
The doctor sent Google a detailed explanation at the end of August. In mid-September, Google indicated it would take no action, and a formal notice sent the same day went unanswered. The court then ruled urgently, without having heard Google beforehand, providing for fines of up to 250,000 euros per violation. Google can still challenge this order.
The court’s reasoning is more noteworthy than the outcome of the dispute. The Aachen court relied on a prior ruling by the Munich Regional Court, which held Google directly responsible for AI Overviews associating two publishers with scams and subscription traps, based on claims absent from all cited sources. Google argued that users could check these sources themselves. The court rejected this argument, stating that the summary read as a standalone statement, and described these outputs as ‘independent, new and substantial statements’ constructed from third-party website content. The Aachen court adopted this vocabulary. Google announced an appeal against the Munich decision, then, according to The Decoder, reached a settlement before a higher court could rule.
A Divergent Reading in Berlin
In June, a Berlin court reached the opposite conclusion in a trademark dispute. A perfume company contested AI Overviews that listed its brands alongside links selling cheaper imitations. The Berlin court considered these summaries to be a new format of search result aggregating information from other sites, not statements by Google. It found no infringement.
The cases do not overlap perfectly. According to the analysis by Duane Forrester, author of the original column, the Berlin summary faithfully reproduced what other sites said, whereas the Munich and Aachen responses established links that no source made. Courts in the same country have nevertheless described the same product in incompatible terms, and no higher court has ruled on the matter.
For the author, the mechanism at play is more mundane than the result. A system that answers a question about a person or company retrieves documents matching the name and related attributes, such as role, institution, city, or a recent event, then generates a single narrative from these elements. When two entities share enough attributes, the system merges their information, which the author calls ‘entity conflation’. In 2024, Microsoft Copilot described a German court reporter as the perpetrator of crimes he had been covering for years, his name appearing next to these cases in his own articles.
