A German court has ruled for the first time that Google bears direct legal responsibility for defamatory content generated by its AI Overviews feature, issuing a preliminary injunction on May 28, 2026, that prohibits the search company from repeating specific false claims about two publishing companies that are part of a Munich-based media group.

The Munich ruling

The Landgericht Munchen I - Munich's regional civil court - issued its judgment on May 28, 2026, under case reference 26 O 869/26. The court's 26th civil chamber, presided over by a senior judge along with two associate judges including one named Muhlbauer, reached its decision following an oral hearing held on April 23, 2026.

The two plaintiffs are a publisher operating 12 branded imprints spanning several content categories, and a subsidiary that publishes books and magazines under the GeraMond brand, concentrating on technology and history. Both companies have their registered seat in Munich. The defendant is Google, which operates google.de in Germany.

The injunction prohibits Google from asserting or disseminating in its AI Overviews - referred to in German as "Ubersicht mit KI" - a series of specific claims about the two companies. These include allegations that the companies engage in or are known for fraudulent schemes or disreputable business practices, that they lure customers into subscription traps by inducing them to sign up for paid subscriptions without their knowledge, that they invoice customers for services such as "company directory listings" or "Premium Gold Packages" following telephone calls that never took place, that they continue demanding payment after customers have already paid, that they frequently change their names and URLs to evade identification, that they fail to unlock paid digital content, and that they are unreachable by phone and ignore written enquiries.

For each violation of the injunction, Google faces a possible fine of up to EUR 250,000 per instance, or a custodial sentence of up to six months enforced against members of its executive management, under the terms set out in section 890 of the German Code of Civil Procedure (ZPO). The total custodial exposure is capped at two years.

Costs were split: according to the court's judgment, the defendant must bear 80 percent of court and legal fees, while each plaintiff bears 10 percent of their own costs.

How the AI Overviews content appeared

The case originated when a search was conducted on January 20, 2026, using the first plaintiff's company name alongside the German word "Betrugsmasche" - meaning fraud scheme. According to the court document, the AI Overviews feature returned a structured response with an affirmative opening line asserting that the publisher was known for disreputable business practices and was frequently perceived as operating a fraud scheme. The response then set out sections labelled "characteristics of the alleged fraud scheme" and a further section headed "what you can do."

A second search on January 26, 2026, returned a similar structured response. A third search on February 10, 2026 - conducted after the companies had already sent a formal legal warning to Google - returned a response that was similar in substance, according to the court record.

The court found, citing the internal structure of those responses, that the content went well beyond the simple display of hyperlinks to third-party web pages. The AI-generated output assembled and synthesised material from multiple external sources, gave it a thematic structure that did not exist in any single underlying source, and presented an independent answer to the query rather than a list of search results. In several instances, the court found that claims in the AI Overview - including specific assertions linking the plaintiffs to other named companies - did not appear in any of the underlying web pages that the feature cited as references.

"Auch ist eine 'Ubersicht mit KI' - anders als die Anzeige von reinen Suchergebnissen in Suchmaschinen - fur die Nutzung des Internets auch keineswegs zwingend erforderlich," the court stated, translating roughly as: an AI Overview is, unlike the display of pure search results in search engines, by no means indispensable for internet use.

The four arguments the court made

Natali Helberger, Distinguished University Professor of Law and Digital Technology at the University of Amsterdam and director of the AI, Media & Democracy Lab, summarised the court's reasoning in a LinkedIn post on June 10, 2026: "The court essentially makes four arguments 1) the presentation of the results, with a summary and structured thematic presentation, which clearly distinguishes it from a search result, 2) the fact that it is Google that offers KI Overviews and has sole control over the technology, 3) a simple search for the name of the company does not produce similar results and 4) the court argues that search engines have an essential function for the existence of the internet by helping to navigate the 'flood of information' online, which is not true for Google Overviews."

Helberger described the ruling as "a preliminary decision, but brings up some very relevant argument for the liability for the output of AI Overviews and similar services."

Each of the four arguments deserves careful examination. The structural distinction is central. Regular search results are a ranked list of hyperlinks with brief textual previews (snippets). An AI Overview does something different: it synthesises, summarises, and structures. The court found that by doing so, Google creates original expression of its own rather than merely pointing to the expression of others.

On the question of control, the court noted that Google itself introduced the AI feature and offers it to users. The algorithms determining what the AI outputs, how it weights sources, and how it structures its answers are entirely under Google's authority. Because only Google can influence those algorithms, the court found that Google must also accept responsibility for their outputs.

On the issue of whether a conventional search would have returned the same content, the court examined the underlying sources that the AI Overview cited as references. None of them, according to the judgment, contained the specific connections that the AI drew between the plaintiffs and other named companies. The AI had generated claims that were novel to its own output.

The fourth argument separates AI Overviews from the long-established legal treatment of search engines. German courts, following Federal Court of Justice (BGH) precedent, have traditionally granted search engines reduced liability on the ground that they are indispensable infrastructure for internet use, making it impractical to impose proactive review obligations. The Munich court declined to extend that reasoning to AI Overviews. The feature is supplementary, the judges found - users can already navigate the "flood of data" using ordinary search results. Presenting this additional layer of synthesised commentary therefore does not attract the same strong public-interest protection that shields conventional search.

Google's defence and why it failed

Google argued that the injunction requests were too vaguely worded, that the AI Overview content did not constitute "assertions" by Google because the company was merely aggregating third-party information, and that the territorial scope should be limited to Germany.

The company also contended that it should be treated as a hosting provider under Articles 4 to 6 of the EU Digital Services Act (Regulation (EU) 2022/2065), which provides liability exemptions for platforms that store and transmit third-party content without active involvement in its creation. Alternatively, Google argued it should be treated as a search engine subject only to a "notice and take down" standard - meaning it should not face liability until it receives specific notice of an obvious legal violation.

The court rejected all these arguments. On the DSA point, the judges found that DSA Article 6(4) explicitly preserves the right of courts to require cessation of violations through civil injunctions, and that in any event Google cannot claim passive hosting status when the content in question was generated by its own AI system. On the notice-and-take-down argument, the court found that the plaintiffs had in fact notified Google twice - first by a formal legal letter dated February 2, 2026, and then again through the online reporting form that Google directed them to use. Despite those notifications, a search on February 10, 2026, still returned substantially similar AI-generated content. Google had therefore had the opportunity to act and had not done so.

The court also engaged directly with Google's argument that attaching links to source pages was sufficient to put users on notice that the content was based on third-party material. The court found this argument unpersuasive. The AI Overview, it said, presents itself as a complete and self-contained answer. Users reading it casually - what German media law calls "Titelseitenleser," roughly "front-page readers" - have no reason to verify each claim against the underlying links, especially when the overview contains no internal cues suggesting it might be unreliable.

The AI Act dimension

The court briefly addressed the EU AI Act (Regulation (EU) 2024/1689), which entered into force progressively from 2024. According to the judgment, the provisions relevant here do not override the civil law claims; they provide only the additional option of lodging a complaint with a market surveillance authority, and do not displace national personality rights remedies. This is a significant procedural clarification for companies considering whether to pursue AI-related claims through regulatory channels or through civil courts.

Why the disclaimer was not enough

Google's AI Overviews carry a label identifying the content as AI-generated. The company argued in proceedings, according to the judgment, that this label should be sufficient to put users on notice that the content might be inaccurate and that they should check the underlying links. The court dismissed this argument by reference to the established German concept of the "Titelseitenleser." The fact that a statement could theoretically be falsified through further research does not release the author from liability for the statement itself - any more than a newspaper headline is protected from defamation claims simply because the full article, further down the page, might contain qualifications.

This reasoning has direct implications for how AI disclaimer practices interact with liability. A label that reads "this content is AI-generated" does not, in the court's view, transfer the verification burden from the publisher to the reader.

The gap in personality-rights protection

One of the more technically intricate passages of the judgment addresses why the usual hierarchy of claims - first check whether someone else is responsible, then hold the platform only if ordinary search logic fails - is not adequate for AI-generated content.

The court reasoned as follows. When an AI Overview contains a false claim, the affected company cannot sue the web pages from which the AI drew its material, because those pages may not have contained the specific claim the AI generated. The claim lives only inside the AI Overview itself. If Google were treated merely as a passive intermediary that faces liability only for obvious violations after receiving notice, and if the false claim is subtle rather than obviously unlawful, then no one is legally responsible for the harm. The affected company would have no effective remedy.

That gap in protection, the court found, is a further reason why AI Overviews must be treated differently from ordinary search results and from autocomplete suggestions.

Context: a growing body of German case law

This ruling does not emerge in a vacuum. A Frankfurt Regional Court decision in September 2025, covered by PPC Land, addressed a similar question when plastic surgeons challenged a false AI Overview about medical procedures. That court dismissed the injunction on the facts - it found the AI content was not actually false when read in full context - but confirmed that Google could in principle be liable for objectively incorrect AI Overview statements. The Munich ruling goes further, actually granting relief.

German businesses' systematic use of Digital Services Act mechanisms to remove online reviews and the Dusseldorf Regional Court's January 2025 platform-liability ruling against Google over review removal both demonstrate that German courts have been developing an increasingly active posture toward platform accountability in the years leading up to this judgment.

In the United States, the pattern of AI defamation claims has been building since 2025. Wolf River Electric sued Google in March 2025 after Gemini fabricated statements about a non-existent attorney general lawsuit against the company. Conservative activist Robby Starbuck filed a $15 million defamation suit against Google in October 2025 over AI-generated false content. Both cases rely on different legal frameworks from the Munich ruling, but the factual pattern - AI-generated false content causing concrete reputational and commercial harm - runs through all of them.

Separately, Google faces an antitrust complaint filed at the European Commission in June 2025 by independent publishers, and a formal EU investigation into AI content practices that opened in December 2025. Those proceedings address different issues - traffic diversion and publisher compensation - but they form part of the same broad regulatory context in which the Munich ruling sits.

What this means for the marketing community

For companies in digital marketing and advertising technology, the Munich ruling introduces a category of legal risk that previously had limited precedent. Businesses that purchase search advertising operate within an environment where the organic search results alongside their ads now include AI-generated summaries. Those summaries can, as this case shows, contain false claims about competitors or third parties, drawn from and attributed to sources that do not actually support those claims.

The ruling is a preliminary injunction - technically a form of emergency relief - and the losing party can appeal. It has not been confirmed by a higher court. The argument that AI-generated content constitutes the operator's own speech, rather than third-party content merely displayed by a passive intermediary, may be tested further. However, the logic the Munich court applied is detailed and draws on established BGH precedent in a methodical way; it is not easily dismissed on appeal simply as novel or unsupported.

More broadly, the ruling raises a question that the EU antitrust investigation and multiple studies on AI Overviews' traffic impact have approached from different angles: what obligations attach to a search feature that generates novel content rather than curating existing content?

The court's answer, at least for now, is that the operator of such a feature bears the same responsibility as any other publisher of original content - and cannot shelter behind the infrastructure protections that have shielded conventional search since the early 2000s.

Timeline

  • January 20, 2026 - First documented AI Overview response to a search combining the plaintiff's company name with the word "Betrugsmasche" (fraud scheme), containing the claims at issue
  • January 26, 2026 - Second AI Overview search returns a similar but distinct structured response
  • February 2, 2026 - Plaintiffs' lawyers send formal cease-and-desist letter to Google; Google's email address for complaints responds the same day saying it cannot process the complaint and redirects to its online reporting form; complaint is resubmitted through that form
  • February 10, 2026 - A third AI Overview search, conducted after the formal notice was sent, returns a substantially similar response
  • February 24, 2026 - Google requests additional information about the complaint
  • April 23, 2026 - Oral hearing at the Landgericht Munchen I
  • May 28, 2026 - Landgericht Munchen I issues judgment (case 26 O 869/26), granting the preliminary injunction against Google and awarding 80 percent of costs against the defendant
  • June 10, 2026 - Natali Helberger, law professor at the University of Amsterdam, publishes analysis of the ruling on LinkedIn, noting its significance for AI Overview liability

Related PPC Land coverage:

Summary

Who: The Landgericht Munchen I (Munich Regional Court I), in proceedings brought by two German publishing companies forming part of a 12-imprint media group based in Munich, against Google, operator of google.de.

What: The court issued a preliminary injunction on May 28, 2026, finding Google directly and primarily liable for false and defamatory claims generated by its AI Overviews ("Ubersicht mit KI") feature during searches combining the plaintiffs' company name with the word "Betrugsmasche" (fraud scheme). The injunction prohibits Google from repeating those specific claims, on pain of fines up to EUR 250,000 per violation or custodial sanctions. Google must bear 80 percent of court and legal costs.

When: The harmful AI Overview responses were first documented on January 20, 2026. Formal notice was sent to Google on February 2, 2026. The oral hearing took place on April 23, 2026. The judgment was issued on May 28, 2026.

Where: Germany. The court established jurisdiction at Munich on the basis that both plaintiff companies are headquartered there and the AI Overview responses were accessible in Germany. The injunction is not limited to German territory, however, as the court applied EU civil jurisdiction rules that allow broader-scope orders.

Why: The court found that AI Overviews constitute original speech by Google rather than a mere display of third-party content. Because the AI independently synthesised, structured, and in some instances fabricated connections between the plaintiffs and other companies - connections not found in any of the cited source pages - Google could not rely on hosting-provider or search-engine immunity. The AI Overview feature is also, the court held, not indispensable to internet use in the way that conventional search is, and therefore does not attract the same strong liability protections that German courts have historically extended to search engines.