A court in Munich has handed down a landmark ruling that could change the rules of the game for AI-powered search services worldwide. For the first time, a search engine company has been held directly liable — on a par with a traditional publisher — for misinformation generated by artificial intelligence.

Google's AI linked the publisher to fraud

The case arose after Munich-based publisher Verlagshaus24 and one of its subsidiaries discovered that Google's AI Overviews feature was portraying them in a highly negative light. According to the source material gathered, the AI system had mixed information about legitimate companies with information about companies engaged in dubious practices, creating connections that did not exist in the original sources. The publisher was consequently and incorrectly associated with fraud, subscription traps, and shady business practices.

The Munich Regional Court heard the case under docket number 26 O 869/26 and issued a preliminary injunction against Google on 28 May 2026. The court also ordered Google to cover 80 percent of the legal costs.

German court: Google is liable for AI lies about businesses - Bilde 1

No longer just a link aggregator

The most legally significant point in the ruling is the distinction the court draws between traditional search results and AI-generated summaries. The court concluded that AI Overviews cannot be equated with ordinary link indexing, but instead constitute Google's own content.

The reasoning is that the AI system reformulates and evaluates results on its own terms, generates independent and substantive claims, and that Google alone controls what is produced. This differs fundamentally from the established EU legal norm, under which providers of traditional search results are generally exempt from liability for claims made in third-party content.

Providers cannot hide behind AI when misrepresentations occur. New and innovative solutions must always comply with the existing legal framework.

It is attorney Bernhard Buchner, who represented the plaintiffs, who is quoted making this statement according to the source material. IT lawyer Stefan Lutz commented in turn that the ruling strengthens the position of companies whose reputations are damaged by AI-generated search results.

Google's defence did not hold up

Google argued, among other things, that users can consult the underlying sources themselves, and that it is widely understood that AI-generated content should not be trusted blindly. Both arguments were rejected by the court, which did not consider the responsibility to lie with the user.

91 percent accurate — but the sources don't check out

An interesting backdrop to the case is the accuracy figures for Google's AI Overviews. An analysis reviewed by the New York Times — and referenced in the source material — indicates that the system, powered by Gemini 3, is correct in approximately 91 percent of cases. Yet the picture is more nuanced than that figure alone suggests.

91%
Correct AI answers (estimate)
50%+
Correct answers without source support

More than half of even the correct answers are not supported by the sources cited in the overview. This raises questions about transparency and traceability that go beyond the accuracy rate alone.

Could set a global standard

Legal observers note that the Munich ruling could potentially have consequences well beyond Europe's borders. The decision is described as one of the first court rulings globally to assign publisher-level liability to an AI-powered search product.

For the first time, a court is holding a search engine company liable as a publisher — not as a neutral link distributor.

The ruling could challenge protective mechanisms that in the United States correspond to Section 230 of the Communications Decency Act — rules that have traditionally shielded platforms from liability for third-party content. The question now being raised is whether the same principle can apply to content that an AI model itself generates, rather than merely distributes.

According to source material from Wired, legal experts believe the decision helps shape a new standard for AI governance that could affect enterprise platforms worldwide, and that it underscores the growing pressure to clarify who bears responsibility when AI systems produce inaccurate, defamatory, or harmful information.

The case has so far been decided as a preliminary injunction. It remains to be seen whether it will be tested in a full trial, and whether other jurisdictions will follow the Munich court's reasoning.