A regional court in Munich issued a preliminary injunction against Google in May 2026, holding that misleading text generated by AI Overviews should be treated as Google's own statement rather than a neutral reflection of web content. On that basis, the court said the feature cannot rely on the traditional liability shield associated with search engines, and a general warning that AI may be inaccurate does not remove responsibility.
The dispute was brought by two German publishers. They argued that AI Overviews displayed clear and affirmative negative claims about their brands in Google Search, including language saying a publisher was known for questionable business practices and was often regarded as a scam. The court highlighted a key detail: some of the contested claims did not appear in the listed search results at all. That meant Google could not defend the summaries as a faithful presentation of third-party sources because the damaging language was not present in those sources.
The court drew a line between search listings and AI-generated assertions
Google argued along a familiar industry line, saying users understand AI output is not always accurate and should verify it independently. The court rejected that premise. In its view, a conventional search engine acts as a guide by showing links, titles, and snippets from third-party websites, while the source of the claim remains external. AI Overviews works differently. It produces an independent and substantive statement directly on the search page, turning the speaker from a third party into Google itself.
That shift became the legal hinge of the case. Once the content is treated as Google's own expression, the responsibility changes as well. The court also dismissed the idea that the publishers should pursue third parties instead. It said only Google can change the underlying algorithm and prevent the same output logic from generating similar statements again.
Failure to correct the output helped trigger the injunction
The publishers had already sent Google a formal notice earlier this year demanding correction of the false output. According to the source material, Google did not take corrective action after receiving that notice. The court then stepped in with a preliminary injunction ordering Google to stop distributing the disputed statements until the main proceedings are resolved and to prevent them from appearing in later AI Overviews results.
A preliminary injunction is not a final ruling, but it shows the court found the publishers' claims credible enough to justify immediate intervention. For Google, the more damaging point was the court's observation that the summaries included statements absent from the underlying search results. That sharply weakens any argument that the company was only passing through information published elsewhere.
The warning extends beyond Google to AI search and chat services
The significance of the ruling reaches past a single defamation dispute. Any service built around AI summaries, rewritten web material, or direct synthesized judgments may face the same question: if an AI system issues a definite negative statement about a person or company, and that statement is false or unsupported by source material, can the provider still avoid liability by saying users should not rely on the output as fact.
At least in Germany, this ruling points to a clear answer. The court said the false outputs were primarily an expression tied to the defendant's commercial activity, and that AI-generated opinions can influence public opinion. The source material adds that the reasoning carries warning value for services such as ChatGPT, Claude, and Perplexity, which also rewrite or summarize information from the web.