The written reasoning of the Munich Regional Court I judgment of May 28, 2026, in case 26 O 869/26, rests on three arguments that together dismantle the liability shield Google has held in Germany since the early 2000s: AI Overviews speak in their own words, they contain claims found in no source, and treating them like search results would leave victims of machine-generated falsehoods without any legal remedy.

The outcome of the case is known. PPC Land reported in June that the 26th Civil Chamber issued a preliminary injunction prohibiting Google from repeating false claims its AI Overviews generated about two German publishing companies, backed by fines of up to EUR 250,000 per violation. The certified copy of the judgment, delivered after an oral hearing on April 23, 2026, now permits a closer reading of how the court reached that result - and where the same reasoning could carry future disputes over generative search.

The stakes of the doctrinal question exceed this single case. For two decades, German courts, following the Federal Court of Justice, treated search operators as indirect interferers: liable only after notification of an obvious infringement, and never obligated to check content proactively. The Munich chamber declined to extend that framework, holding that the display of the contested "results with AI" is "not a mere display of search results, but rather its own content attributable to it." Everything else in the judgment flows from that classification.

case 26 O 869/26
1781123511842.pdf
1 MB
download-circle

The first ground: a voice of its own

According to the judgment, the decisive difference between a results page and an AI Overview begins with language. Search results appear as links or snippets. The overviews at issue summarized and presented the results "in the site's own words and according to its own structure," opening with an affirmation of the query itself: "Yes, Verlagshaus24 (GeraMont Verlag) is known for dubious business practices..." in one version, and "Yes, there are indications of scams and dubious practices in connection with Verlagshaus24..." in the other.

The court found that this affirmative opening "linguistically goes beyond the mere presentation of links." The texts then imposed a thematic architecture found in none of the cited sources: an introductory summary, a compilation under headings the judgment renders as "Characteristics of the alleged scam" and "Typical Scams," and closing recommendations under "What You Can Do" and "What You Should Do." All of this, the chamber wrote, demonstrates an independent, content-based processing of the search results by the AI offered by the respondent. Because Google introduced the system and alone controls its algorithms, it must answer for what the system produces.

The second ground: claims that exist nowhere else

The argument the court described as weighing "above all" concerns fabrication. According to the judgment, none of the links displayed alongside the overviews established any connection between the publishing companies and the third-party firms the AI associated with them, nor did any source claim frequent name changes between the entities. The statements originated in the overview itself.

The sourcing detail is stark. Both overviews cited, in first position, an article by the law firm Loschelder Leisenberg that "contains no reference whatsoever to the plaintiffs." An affidavit from the Association of German Publishers and Booksellers, submitted as Exhibit ASt 14, established that no connection exists between the publishers and the companies named in the AI text. Google did not contest this with substantiated argument. On that record, the court concluded the overview constitutes "a separate statement generated by the AI provided by the respondent, for which the respondent, as the provider (also within the meaning of Art. 3(3) of the AI Regulation), must accept responsibility."

The reference to Article 3(3) of the EU AI Act is a quiet but notable move. The court did not apply the regulation as a basis for the claim - it expressly held the AI Act offers only a complaint route to market surveillance authorities - yet it borrowed the regulation's provider concept to anchor responsibility for system output in national civil law.

The third ground: a gap no one else can fill

The chamber then asked what would happen if AI-generated statements received the notice-and-takedown treatment developed for linked content. Its answer supplies the judgment's most consequential passage. Affected parties cannot sue the operators of the source websites, because those third parties never made the fabricated statements; the claims were "first compiled and disseminated by the search engine operator" in the overview. If the operator then owed nothing beyond removing obvious violations, victims of statements that are unlawful only after thorough examination "would then have no adequate opportunity to obtain legal protection."

That protection gap, according to the court, is itself a reason why the precedents on search engines and autocomplete cannot be transferred to AI Overviews.

Distinguishing two decades of BGH precedent

Google built its defense on two Federal Court of Justice decisions: the 2018 search engine ruling in case VI ZR 489/16 and the 2013 autocomplete ruling in case VI ZR 269/12. The Munich judgment works through both.

The 2018 decision shielded search operators because a proactive review duty "would conflict with the purpose and functioning of a search engine," and because, without search, the internet "would be unusable for individuals due to the overwhelming flood of data." The Munich chamber accepted that premise and denied its application. An AI-powered overview, the court held, "is by no means essential for using the internet," since displaying links already makes the flood of data manageable. The feature instead "structures and evaluates data according to a system not immediately recognizable to the user" and thereby channels the answer. Convenience, in the court's analysis, does not earn the constitutional solicitude granted to necessity.

The court also disposed of the practical argument behind the 2018 ruling: that a search operator has no relationship with the authors of indexed pages and cannot investigate their claims. Google faces no such obstacle here. The company "is certainly capable of verifying the content of the AI Overview against the sources on which it is based, even without contacting the third parties." According to the judgment, Google failed to do so even after the publishers notified it - by attorney letter on February 2, 2026, by email the same day, and again through Google's own online form after the company replied that it could not process the emailed complaint.

The autocomplete precedent fared no better. In 2013, the BGH found operators responsible in principle for search suggestions but limited duties to cases of notified infringement. Munich distinguished the products: at issue is neither a query suggestion nor a list of links but "an independently generated statement that, based on the weighting of the search results and an evaluation of the content, formulates its own answer (not a question) and offers it to the user."

The chamber left open whether a review duty triggered only by notification might suffice for AI output, because the publishers had undisputedly notified Google through two channels. It rejected, however, any limitation to obvious violations, given that the system's own sources permit verification.

The front-page reader meets the AI answer

During the oral hearing, Google argued that informed users know AI-generated information should not be blindly trusted and can check the attached links. The court answered with an established doctrine of German press law. Casual readers - the judgment invokes the figure of the front-page reader, developed by the Federal Constitutional Court in 1998 - may take a self-contained statement at face value, and the possibility of refuting a statement through further research does not relieve the speaker of liability for it.

Applied to the product: the overview "is comprehensible in and of itself, contains a self-contained statement with independently understandable content, and makes no reference to other possible interpretations or even to any unreliability of the content, so that users generally have no reason to review the displayed answer to the search query." The court added a pointed observation about product logic. If overviews were generally recognized as unreliable and every link required independent verification, this "would significantly diminish the very utility of the feature cited by the respondent."

Opinion built on invented facts

The judgment's second half classifies each contested statement as fact or opinion, and the taxonomy explains both what Google lost and what it won.

The headline assertion - that the publisher is known for dubious business practices and often perceived as a scam - qualified as an expression of opinion, because factual and evaluative elements were inseparably intertwined. Under German constitutional law, opinions need not be well-founded, substantiated or convincing. The protection collapsed anyway, for a reason the court located in Federal Constitutional Court case law: the weighing must consider whether an opinion rests on true facts or on "an assessment arbitrarily plucked out of thin air." Here the opinion rested primarily on the alleged connection to subscription-trap companies, a connection the affidavit disproved. An opinion built on a false connecting fact loses the balancing.

The court added a consideration specific to machine output. The opinion "was primarily generated by AI, meaning it is not an expression of a conviction held by the persons making the statement, but rather the result of an algorithm." Offering AI-supported search, the chamber found, is primarily commercial activity and only secondarily participation in public discourse - a framing that lowered the weight of Google's expression interest in the balance.

Statements open to proof - the company associations, the billing after telephone calls that never took place, the failure to unlock paid content, the unreachability - were treated as factual assertions. Because they damage reputation, the burden of establishing truth fell on Google under the principles of Sections 186 and 187 of the German Criminal Code. Customer reviews on rating portals did not carry that burden: according to the judgment, the reviews concerned dissatisfaction with service and order processing, not unwitting subscription sign-ups, and review platforms do not verify what third parties post. Nor had Google shown that it, or the AI it employs, had subjected the statements to any review at all.

Where the plaintiffs lost

Two claims failed, and the failures trace the limits of the doctrine. The allegation of cooperation with a debt collection agency was, in the court's view, not damaging enough to shift the burden of proof to Google; the publishers would have had to establish its untruth and did not. And the asserted statement that the plaintiffs "sell subscriptions and collect payments in an unethical manner" appears in neither overview. At most readers might infer it, and an inference formed by readers cannot be prohibited unless it imposes itself as an inevitable factual claim. The cost split reflects the outcome: each plaintiff bears 10 percent of court fees, Google 80 percent.

Google's final argument - that a machine learning revision had removed the texts, eliminating repetition risk - also failed. Without a penalty-backed cease-and-desist declaration, and given that recurrence "cannot be reliably ruled out" under the algorithms in use, the risk persisted. The injunction, the court held, is not limited to German territory, because Article 36(1) of the Brussels I Regulation extends recognition of the decision across the regulation's scope.

Why this matters for the marketing community

The reasoning arrived in a legal environment that had been circling the question. A Frankfurt Regional Court decision of September 10, 2025 accepted in principle that an objectively incorrect AI Overview could constitute unlawful obstruction, but dismissed the case because the summary proved ultimately not false. Munich supplies what Frankfurt lacked: a documented falsehood, and a full doctrinal pathway from fabricated output to enforceable injunction.

Commentary from the advertising industry read the judgment through the economics of zero-click search. "This judgment hits the nail on the head," wrote Daniel Rijo, a programmatic marketing professional at Havas Media Germany, on LinkedIn, noting far-reaching implications including for competition law. "Economically, technically and legally, this is compelling: if you synthesise third-party content and add some hallucinations to give your comprehensive own answer, thereby eliminating incentives to click through to the sources, you cannot argue that you are merely a neutral host provider. You are accountable for your answer, even if it is wrong."

The click-through dimension he references is quantified. Research across 300,000 keywords found organic clicks to top-ranking pages fall 34.5 percent when AI Overviews appear, and the Independent Publishers Alliance complaint of June 30, 2025 put that traffic loss before the European Commission. The Munich reasoning connects the two grievances: the same synthesis that removes the incentive to click is what converts Google from intermediary to author. The European Commission is separately preparing a record Digital Markets Act fine that includes concerns about AI Overviews self-preferencing, and the Court of Justice confirmed the 4.1 billion euro Android fine on July 2, 2026, so the liability finding lands on a company already defending its search model on several European fronts simultaneously.

For businesses of any size, the doctrinal shift changes what a false AI Overview means. It is no longer third-party content awaiting discretionary review but the operator's own speech, contestable under corporate personality rights with per-violation penalties attached. The judgment remains preliminary relief and can be appealed. Its architecture, assembled step by step from Federal Court of Justice and Federal Constitutional Court precedent, was plainly built with that appeal in mind.

Timeline

  • January 20, 2026: A search combining the publisher's name with the term "Betrugsmasche" returns an AI Overview affirming the company is known for dubious business practices
  • January 26, 2026: A second search returns a similar overview; both cite a law firm article containing no reference to the plaintiffs
  • February 2, 2026: The publishers' attorney sends a cease-and-desist letter and email; Google replies the same day that it cannot process the complaint and refers to its online form, through which the complaint is resubmitted
  • February 10, 2026: A further search, after the warning letter, still yields a summary text similar in key respects
  • February 24, 2026: Google requests further information regarding the inquiry
  • April 23, 2026: The Munich Regional Court I holds the oral hearing
  • May 28, 2026: The 26th Civil Chamber issues its judgment in case 26 O 869/26, granting most of the injunction and assigning Google 80 percent of costs

Summary

Who: The 26th Civil Chamber of the Munich Regional Court I, ruling for two German publishing companies - a house bundling 12 publishing brands and its subsidiary operating the GeraMond imprint - against Google as operator of google.de.

What: The written judgment classifies AI Overviews as Google's own statements on three grounds: they summarize results in their own words and structure, they contain claims found in no source, and shielding them would leave affected parties without legal protection. The classification strips Google of the hosting exemption under the Digital Services Act and the notice-and-takedown privilege for search engines, exposing the company to fines of up to EUR 250,000 per repeated violation. Two contested statements survived, and Google bears 80 percent of costs.

When: The judgment was issued on May 28, 2026, following an oral hearing on April 23, 2026, over AI Overviews displayed between January 20 and February 10, 2026.

Where: Munich, Germany, case number 26 O 869/26, with effect extending beyond German territory under Article 36(1) of the Brussels I Regulation.

Why: The court found the overviews fabricated connections between the publishers and fraud-linked companies, cited a source that never mentioned the plaintiffs, presented self-contained answers that front-page readers have no reason to verify, and rested a damaging opinion on untrue connecting facts - making the texts Google's own speech under German corporate personality rights law.