Is Google actually a publisher when the search engine’s answers are written by AI rather than simply pulled from external websites? In Germany, there is now a preliminary answer to this question.

The Regional Court of Munich I (Landgericht München I) has issued a preliminary injunction prohibiting Google from spreading false claims about two publishing houses in its AI Overviews – under threat of a fine of up to 250,000 euros. The reasoning is remarkable: AI-generated search summaries, the court found, are Google’s own content, not mere search results.

For decades, search engines have enjoyed a comparatively comfortable liability regime: anyone who merely links to third-party websites is, in most jurisdictions, only liable once they have been notified of an obvious legal violation and fail to act. This privileged status as an “indirect interferer” or neutral intermediary – enshrined in the EU, among other places, in the Digital Services Act (DSA) – did not help Google in the proceedings before the Munich court.

In a ruling dated May 28, 2026 (case no. 26 O 869/26), the court prohibited the company, in expedited proceedings, from disseminating several untrue factual claims about two Munich-based publishing companies in its AI Overviews – known in German as “Übersicht mit KI.” Violations carry a fine of up to 250,000 euros or detention. It is considered the first court decision of its kind concerning Google’s AI search feature, which the company has been rolling out worldwide since 2024.

What the Case Was About

The dispute was triggered by AI-generated texts that Google displayed prominently above its classic search results. When users searched for the name of the plaintiff companies – a publishing house and its subsidiary – in combination with terms such as “scam,” the AI Overviews produced their own summaries. Among other things, these claimed that the companies were known for dubious business practices, lured customers into “subscription traps,” ignored written inquiries, and were linked to shady firms.

The problem: none of these statements appeared in any of the linked sources. According to the court’s findings, the AI had mixed information about other, genuinely questionable companies with the plaintiffs and constructed connections that did not exist – a textbook case of AI hallucination. The publishers saw this as a violation of their corporate personality rights. Google initially did not respond to a cease-and-desist letter, prompting the companies to go to court.

“Independent, New Statements” Rather Than Search Results

Google defended itself with its usual argument: it merely displays information from third-party websites in an automated fashion, does not adopt this information as its own, and is therefore liable, at most, as an intermediary. The Munich court did not follow this view – and classified Google as a direct interferer, directly responsible for the content.

The reasoning: AI Overviews go far beyond merely linking to websites. By summarizing, structuring, and connecting information from various sources, the AI makes independent, new, and substantive statements – in some cases statements that do not appear in the underlying sources at all. Because Google offers the feature itself and controls the algorithms, the company must accept the generated content as its own. From the perspective of average users, the AI summary appears to be direct information from Google, not a mere pass-through of third-party content. Google therefore cannot invoke the liability privileges for hosting providers and classic search engines, such as those under the DSA.

The court also rejected another line of defense: the argument that users could verify the AI answers themselves using the linked sources does not exonerate Google. On the contrary – if users had to double-check every statement anyway, the practical value of the summaries would be largely lost. Moreover, the court found that monitoring is entirely reasonable for Google, at least by comparing its own AI statements against the sources used.

A smaller part of the claim was dismissed: Google may continue to state that the plaintiffs work with a debt collection agency, because the companies could not sufficiently substantiate that this specific claim was untrue. Nevertheless, Google must bear 80 percent of the costs of the proceedings.

Publisher or Platform? The Real Fundamental Question

The decision touches on a debate as old as the commercial internet: are tech platforms neutral conduits for third-party content – or publishers that must answer for what they publish? For classic search results, social networks, and hosting services, lawmakers and courts have largely answered this question in favor of the platforms over the years.

The Munich court now draws a line: as soon as a system formulates its own coherent text from third-party sources, it functionally resembles an editorial product more than an act of intermediation. Also notable is the chamber’s assessment in terms of fundamental rights: AI-generated statements are the expression of an algorithm, not of a conviction someone has formed – and therefore enjoy weaker protection than human expressions of opinion. In the balancing of rights, the protection of the publishers’ corporate personality rights outweighed Google’s interest in freedom of expression and commercial activity.

However, the ruling does not formally classify Google as a “publisher” in the press-law sense. It is also a preliminary injunction issued by a regional court, not a decision by Germany’s highest courts. The ruling deliberately departs from the Federal Court of Justice’s existing case law on search engines, and an appeal is considered likely. The contrast with a September 2025 decision by the Regional Court of Frankfurt is also interesting: there, the chamber held that liability for AI summaries is not ruled out in principle, but dismissed the specific injunction request. Munich now goes a step further.

What This Means for Google and Other AI Providers

If the Munich line of reasoning prevails, Google faces a structural dilemma. The company would either have to ensure that the statements in AI Overviews and AI Mode are factually correct – something that can hardly be fully guaranteed given the current state of generative AI and its tendency to hallucinate – or tie its AI answers much more closely to the linked sources. The latter could, paradoxically, benefit precisely those publishers who have been complaining about losses in reach and traffic since the rollout of AI Overviews, because their content is summarized rather than linked.

Beyond the individual case, the question is how far the logic of the ruling extends. The argument – whoever independently generates content is directly liable for it – could in principle also be applied to AI chatbots such as ChatGPT, Gemini, or Claude, whose answers are likewise independently formulated texts. In any case, companies and individuals about whom AI systems make false claims now have a first reference case from Munich showing that legal action can succeed.