Munich Court: Google Can Be Liable for False AI Overviews

The Munich Regional Court issued a preliminary injunction finding Google can be held directly liable for defamatory statements produced by its AI Overviews. The order followed complaints from two German publishers who were falsely described as involved in scams and “dubious business practices.”

The publishers found that AI Overviews shown at the top of search results made claims about their businesses that the linked articles did not support. After the publishers sent cease‑and‑desist letters, similar statements reappeared. The court cited Google’s failure to promptly prevent those repetitions when it granted the injunction. The ruling is preliminary and may be appealed.

The court distinguished AI Overviews from standard search results, describing the summaries as “independent, new, and substantive statements.” The judges wrote that the overviews summarize, rephrase and sometimes introduce facts, and that the models and logic that produce them are controlled by Google.

Because only Google can change the models and the system’s logic, the court concluded the company can be required to stop the system from repeating falsehoods. The injunction requires Google to prevent the specific defamatory claims from appearing again while the case proceeds.

Google’s AI Overviews are powered by the company’s Gemini model. Studies of user behavior show that when an AI summary appears on a search page, fewer users click through to standard results. Separate reviews of the feature find most summaries are accurate but that a substantial share are “ungrounded,” meaning the links provided do not fully support the statements the overview makes. Given the volume of searches Google handles, even a small share of errors can produce a large number of incorrect answers.

The court noted that legal protections for search engines, which generally treat search results as channels for third‑party content, do not automatically apply when a platform generates and publishes new content. The decision draws a legal distinction between listing links and creating narrative summaries that present factual claims.

Legal analysts view the ruling as potentially relevant in other jurisdictions because it focuses on the design and operation of automated summaries and a platform’s ability to prevent repeated false statements rather than the technical origin of an error.