A German Court Just Ruled That AI Overviews Are Google's Own Words—and Made the Company Liable for Every False Statement

The court found that Google's AI had confused the plaintiffs with entirely different companies that actually had problematic histories. Crucially, the defamatory connections appeared in none of the linked sources—the AI invented them by blending and reinterpreting information from unrelated third parties . The publishers first sent cease-and-desist requests to Google in early 2026, but the company did not remove or correct the AI Overviews, prompting the lawsuit .

The Court's Classification: Why AI Overviews Are Not Search Results

The central question was whether Google's AI Overviews function as traditional search results—which merely index and link to third-party content—or as something different. The court ruled that they are fundamentally different.

AI Overviews were classified as Google's own original content, producing "independent, new, and substantive statements" that do not simply reproduce or aggregate third-party material . The court reasoned that because Google builds the AI, offers it to users, and owns the algorithms that generate responses, it owns the content produced. "Google built the AI, Google offered it to users, so Google owns what it produces," the ruling stated .

This distinction is legally decisive. Under German and European law, traditional search engines enjoy limited liability (Störerhaftung) when displaying third-party content. But the court held these protections disappear when the platform generates new, self-contained assertions—even if those assertions are based on an underlying corpus of third-party data .

How the Court Dismantled Google's Defense

Google's primary defense was that AI Overviews function like conventional search results: the company argued it merely provides links to sources, and users are expected to verify claims by clicking through to the original pages.

The court rejected this on multiple grounds.

First, the design of AI Overviews contradicts the argument. The feature displays a prominent answer box above all other results—a presentation that signals completeness and authority. Requiring users to fact-check every AI-generated claim by opening underlying links would defeat the purpose of the feature and is not a reasonable expectation .

Second, the AI Overview does not neutrally surface existing statements; it creates new ones. The defamatory claims about the publishers "appeared in none of the linked sources," and in some instances made assertions that "were not even mentioned in the search results at all" . These were found to be Google's own statements, not anyone else's.

Third, existing legal precedents shielding search engines from liability for third-party content simply do not apply to AI Overviews. The court drew a sharp line between the traditional "ten blue links" and AI-generated summaries, ruling that earlier safe-harbor doctrines cannot be stretched to cover generative AI outputs .

What the Injunction Requires

The preliminary injunction orders Google to:

• Cease and desist from repeating the specific false factual claims about the two Munich publishers in AI Overview responses .
• Remove the existing defamatory AI Overviews containing those false statements .

The order is enforceable under threat of court-ordered fines (Ordnungsgeld) of up to €250,000—or even custodial penalties for Google's legal representatives in Germany—for each violation .

It is important to note that this is a preliminary injunction (einstweilige Verfügung), issued through a fast-track procedure rather than a full trial on the merits. A final ruling in the main proceedings is still pending . Its direct legal force is limited to Germany and to the specific parties involved, though its reasoning is expected to influence courts across the EU .

Broader Implications for AI-Generated Search Content in Europe

The ruling has immediate implications for platforms deploying AI-generated search and answer features, as well as for the broader European regulatory landscape.

End of safe harbor for AI search features. The decision signals that the intermediary-liability protections in the EU's e-Commerce Directive and the Digital Services Act likely do not automatically extend to AI-generated summaries. Courts may treat such features as the platform's own content, triggering direct liability for defamation, false assertions of fact, and unfair competition .

Platforms as publishers. By classifying Google as a "direct infringer" (unmittelbarer Störer) rather than a neutral intermediary, the ruling sets a precedent that any company shipping AI-generated search content can be held directly responsible for its accuracy. This logic could apply to Microsoft's Bing Copilot, OpenAI's web-connected features, and other AI search tools operating in Europe .

Publishers gain new leverage. Previously, a publisher or individual harmed by AI-generated misinformation often had to trace the claim to an upstream source and pursue that party. This ruling creates a more direct route: holding the AI platform itself accountable for what its models produce. It strengthens the position of publishers and content creators whose reputations can be damaged by AI hallucinations .

Operational pressure for technical safeguards. The ruling effectively requires Google and similar providers to implement content filters, fact-checking pipelines, and rapid takedown mechanisms for AI-generated search responses. In practice, platforms now face legal consequences for each false factual statement their AI generates, raising the operational stakes considerably .

Courts ahead of regulators. The decision arrives as European regulators are finalizing enforcement of the EU AI Act and the Digital Services Act, placing courts ahead of the legislative process in defining liability for generative AI outputs. It could encourage a wave of copycat litigation across EU member states, potentially forcing regulators and lawmakers to accelerate rulemaking .

The Munich ruling is preliminary, and its ultimate scope will depend on the final judgment and any appeals. But its core finding—that AI-generated summaries are the platform's own speech—represents a significant legal shift. For companies building AI into search, the message is clear: when an AI model generates factual claims, the platform behind it may have to answer for every word.