Two German Publishers v. Google — Landgericht München Holds Google Directly Liable for False AI Overview Statements

Background

Two Munich-based publishing companies discovered that Google’s AI Overviews — the AI-generated summaries that appear at the top of Google Search results — were falsely associating them with scams, subscription traps, and “dubious business practices.” For certain search queries, the AI incorrectly mixed up information about genuinely disreputable companies with the plaintiffs, drawing connections that appeared in none of the underlying source links the AI had supposedly synthesized. The publishers sent Google a cease-and-desist letter (German: Abmahnung), but Google did not promptly correct the outputs. The publishers then filed an urgent application for a preliminary injunction (einstweilige Verfügung) in the Regional Court of Munich.

The case required the court to decide whether Google’s AI Overviews are protected by the host-provider defenses that traditionally shield search engines from liability for third-party content, or whether those defenses do not apply to AI-generated summaries. Traditional German press and internet law had shielded search engines on the theory that they are passive conduits of information created by others. AI Overviews present a fundamentally different case: the AI does not merely index or display third-party content, but generates new text by evaluating, paraphrasing, and combining information from multiple sources.

The Court’s Holding

The court granted the preliminary injunction and held that Google bears direct liability (German: Täterhaftung, as distinct from secondary or hosting liability) for the false AI-generated statements. The key finding: AI Overviews constitute “independent, new, and substantive statements” by Google itself, not mere hyperlinks to or displays of third-party content. Because the AI synthesizes and paraphrases source material into entirely new text that users read and rely on “without needing to consult the linked sources,” the court rejected the argument that this is fundamentally similar to a traditional search index.

The court also rejected Google’s argument that users should independently verify AI-generated search summaries. The court found that the format and placement of AI Overviews — prominently displayed, presented as authoritative responses — creates a reasonable user expectation of accuracy. Requiring users to fact-check every AI Overview would render the feature meaningless. The court further held that the existing safe-harbor protections for hosting providers under German law and the Digital Services Act (Digitale-Dienste-Gesetz) do not extend to AI-generated summaries, which are “algorithmic editorial outputs” rather than passively hosted content. The order required Google to cease displaying the specific false AI Overview statements going forward.

Key Takeaways

• AI-generated content may be the speaker’s own speech. Courts are beginning to distinguish between search engines that index and link (protected intermediary) and AI systems that synthesize new text (direct publisher). The same platform can attract both treatment for different features.
• Promptness in correcting AI errors matters. Google’s failure to correct the AI Overviews after receiving the Abmahnung was a factor. Companies deploying AI-generated public-facing content may need rapid correction workflows comparable to editorial operations.
• Host-provider safe harbors may not cover generative AI outputs. The DSA and national hosting liability laws were written for platforms hosting user-generated content, not for AI systems generating their own content. This court found the fit too poor to extend the protection.
• This is a preliminary injunction, not a final judgment. The ruling sets out the court’s preliminary view; merits proceedings (Hauptsacheverfahren) will follow, and Google can and likely will appeal or challenge the findings.
• Potential global ripple effects. Germany is an influential EU jurisdiction in tech-platform law. A final judgment adopting similar reasoning could influence how courts across the EU — and potentially beyond — approach AI-generated content liability.

Why It Matters

Google AI Overviews appear in billions of searches every day. If AI-generated search summaries are treated as the platform’s own statements for defamation and press-liability purposes, the operational implications for search engines are enormous: they would face liability exposure for every AI-generated summary containing a false factual claim. That is a qualitatively different risk profile from traditional search indexing, where liability is usually limited to cases where the platform ignores clear notice of illegal content.

For IP and AI practitioners, this ruling is a data point in the emerging global debate over where AI-generated content fits in the existing frameworks built for human speakers and passive intermediaries. The question matters well beyond defamation: similar logic could eventually be applied to AI tools that generate content touching copyright (reproducing protected expression), trademark (using marks in generated text), or right of publicity (synthesizing a public figure’s image or voice). The Landgericht München’s analysis — that AI synthesis crosses a threshold from passive intermediation into active authorship — is one early answer to that question.