German Court Ruling: AI Search Overviews Stripped of Search Engine Liability Shield

German Court Ruling: AI Search Overviews Stripped of Search Engine Liability Shield

A landmark ruling by the Regional Court of Munich (Case No. 26 O 869/26, handed down May 28, 2026) has established a critical legal precedent: Google is directly liable for false claims and defamatory statements generated by its AI Overviews¹. This ruling strips AI-generated search overviews of the traditional liability shields enjoyed by classic search engines, treating synthesized AI output as Google's own published content. Google officially appealed the ruling on June 12, 2026, attempting to frame it as a "narrow error case," but legal scholars and subsequent rulings indicate a systemic shift in European jurisprudence.

The Core Legal Shift: Direct vs. Indirect Liability

Under German law, search engines that merely index and surface third-party links have traditionally been protected under the principles of indirect interferer liability (mittelbare Störerhaftung). Under this model, a search operator is only liable for defamatory or unlawful content once it has been explicitly notified of a manifest infringement and fails to act (a notice-and-takedown regime).

The Munich Regional Court shattered this shield by placing AI Overviews under direct interferer liability (unmittelbare Störerhaftung). The court's reasoning rests on several key distinctions:

1. Authorship vs. Retrieval (Zu-Eigen-Machen): An AI overview does not merely list links or display short, passive previews (snippets). Instead, the generative model independently processes, weight-evaluates, and restructures search results into a new, self-contained narrative in its own words. By doing so, Google "makes the content its own" (Zu-Eigen-Machen).
2. The Indispensability Test: Classic search engines are granted liability privileges because they are practically essential for making the overwhelming flood of internet data usable. The court ruled that AI Overviews enjoy no such necessity; the internet remains perfectly usable without them. Because the feature is voluntary and non-essential, the search-engine privilege falls away².
3. Rejection of the "User Duty to Verify": The court rejected Google's defense that users are responsible for fact-checking the AI's claims by clicking the cited source links. Because the AI's output reads as a self-contained, authoritative answer with no visible warning of unreliability, a reasonable average user ("front-page reader") has no reason to verify the statements.
4. Failure of the "AI Error" Defense: The court made it clear that "AI may make mistakes" is not a legally valid excuse. Since Google introduced the AI and controls its algorithms, it must answer for every word it generates.

The Factual Infringement: Verlagshaus24 vs. Google

The case was brought by two Munich-based publishers, including Verlagshaus24 (publishing under the imprint "GeraMond" for technology and history). In January 2026, search queries combining the publisher's name with terms like "fraud scheme" (suggested by Google's own autocomplete function) displayed an "Overview with AI" that confidently asserted:

"Yes, Verlagshaus24 is known for dubious business practices and is often perceived as a scam..."

The AI overview proceeded to list "typical fraud schemes," falsely claiming the publishers lured customers into "subscription traps," made fake phone calls, changed names and URLs to evade identification, and refused to unlock paid digital content.

In reality, the AI had hallucinated these connections by mixing up information about completely unrelated, sketchy companies with the plaintiffs. The primary source cited by the AI was a warning notice published by a law firm, which did not mention Verlagshaus24 at all. Even though the publishers sent Google a cease-and-desist letter, Google failed to respond appropriately, leading the court to issue a broad injunction and order Google to pay 80% of the legal costs.

Systemic Convergence in European Jurisprudence

The Munich ruling is not an isolated event. Legal analyses, including a review by the University of Oxford, highlight a rapid convergence across different branches of German and European law toward holding AI operators directly liable:

• LG Frankfurt am Main (Case No. 2-06 O 271/25, Sept 10, 2025): Admitted the principle of direct liability for inaccurate AI search summaries on competition grounds.
• LG Hamburg (Case No. 324 O 461/25, Sept 23, 2025): Applied personality rights to hold an operator liable for an AI chatbot's fabricated statements.
• OLG Hamm (Case No. 4 UKl 3/25, May 12, 2026): Held a medical practice fully liable under unfair-competition law for its chatbot's false claims of non-existent specialist titles, admitting an appeal to the Federal Court of Justice.

This legal shift transforms AI hallucinations from a minor technical "bug" into a massive corporate governance and compliance risk. Any enterprise deploying generative systems to summarize web data or interact with the public now legally "owns" the output of those systems.