A German court has ordered Google to stop publishing AI Mode answers that mixed up a local doctor with another man who faced real legal proceedings. The facts were true. They just belonged to someone else.
What happened in Aachen
In a decision dated September 24, 2026, the Regional Court of Aachen acted on an emergency basis after the doctor warned Google in late August and Google said in mid-September it would take no action. The order carries fines of up to €250,000 per violation. Google can still challenge it.
The AI answer surfaced when people searched why the doctor, a former chief physician and professor of neuroradiology, had left his post. AI Mode offered a serious criminal allegation, and follow-up searches about suspension or court proceedings produced detailed accounts. Those proceedings were real, but involved a different former university professor.
Why this ruling matters
Aachen did not invent its reasoning. It adopted language from a Munich court that in May held Google directly liable for AI Overviews that linked two publishers to fraud claims absent from the cited sources. The Munich court described the AI output as “independent, new, and substantive statements” and rejected Google’s argument that users could simply check the links.
Platforms point to source links, but Pew Research found users clicked a link inside the AI summary on only 1% of searches where one appeared. That is exactly why courts read the answer as the product.
But a Berlin court went the other way in June, treating AI Overviews as a search-result format rather than Google speaking for itself. No higher court has settled the split. For marketers, that means the rules on the surface answering customer questions are still forming—and can vary by courtroom.
The real mechanism: entity conflation
The failure is not hallucinated facts. It is entity conflation: when two people or companies share enough attributes, the system merges them into a single confident answer. In Aachen, both men were former university professors leaving a post. In 2024, Microsoft’s Copilot reportedly described a court reporter as the perpetrator of crimes he had covered because his name appeared next to those crimes in his own bylines.
This exposure is not limited to doctors. Similar names in the same industry, franchise locations confused with the parent brand, and executives who share a name with someone in the same field are all candidates.
What marketers should do now
You cannot wait for the courts. Your brand is already appearing in AI answers, so focus on identity hygiene:
- Audit on a schedule: query your brand, executives and products with role, location and event terms such as departure, lawsuit or acquisition.
- Publish disambiguating facts: state founding details, leadership, locations, and what you do not do in plain first-party sentences.
- Add machine-readable identity: use Schema.org Organization and Person markup with sameAs links to authoritative profiles.
- Close gaps yourself: when a leadership change or closure happens, publish your own explanation before the AI reaches for a neighbor’s.
- Document every wrong answer: keep the exact query, answer text, date, cited sources and your correction request.
The safest entity in AI search is the one that is hardest to confuse with someone else. Verified first-party facts are no longer just a ranking signal—they are identity protection.
Source: Search Engine Journal
