AI Overview wrong? The fault lies with Google, according to a German court

The Regional Court of Munich, Bavaria has issued a ruling that draws a clear line in the legal responsibility for content generated by artificial intelligence.

With an emergency order tied to case number 26 O 869/26, the court has established that Google is directly responsible for the statements made within its AI Overview.

The decision comes after the multinational’s search functionality erroneously linked two German publishing houses to improper business practices, scams and subscription-related tricks, mixing information related to other companies and creating links entirely absent from the cited sources.

Google is responsible for its AI’s responses, the verdict

The German judges rejected the extension of jurisprudential protections previously applied to conventional search engines. According to the court, the traditional indexing systems merely render third-party content discoverable, at most indirect liability for the provider.

Conversely, the AI-generated summaries rework and evaluate data according to an autonomous structure and their own words, forming novel and independent assertions. Google controls the algorithm and the technical architecture of the tool, consequently the authorship of the statements belongs to it exclusively.

During the hearing, the defense of Mountain View’s company argued that users are aware of the limits of the technology and can verify the reliability of the summaries by consulting the sources.

The court deemed this objection irrelevant, noting that the theoretical possibility of refuting a statement through subsequent verifications does not absolve the author from civil liability.

Such an editorial dynamic has been linked to the right to press, where an editor is responsible for short texts or self-contained headlines even if the reader chooses not to read the full article.

Moreover, the judges noted a lack of protection for the injured parties, because the defamatory statements were not present on the third-party sites used as sources, effectively preventing any legal action against the owners of those domains.

Constitutionally, the court clarified that the opinions generated by an algorithm do not reflect personal beliefs but commercial dynamics, and therefore enjoy lower protection in terms of freedom of expression than the personality rights of the injured parties.

The consequences

The ruling orders Google to pay 80% of the legal costs and bans reiterating the specific accusations. The issue takes on systemic dimensions when considering the search engine’s global traffic volumes.

An independent analysis conducted by the startup Oumi found that the model Gemini 3 responds correctly in 91% of the cases examined. Although this percentage may seem high for everyday generic use, applied to Google’s scale it translates into millions of wrong answers every hour.

The study also highlights that 56% of correct answers provided by the model do not have verifiable corroboration in the linked sources.

The Bavarian court’s decision underscores how the creation of content not backed by facts exposes technology providers to serious legal risks, a precedent that could rapidly extend to other entities in the sector such as OpenAI, Anthropic and Perplexity.