German Court Rules Google Liable for AI Overview Errors
German Court Holds Google Directly Liable for False AI Overview Statements
A Munich court has issued one of the most consequential rulings in the short history of AI-powered search, declaring that Google is directly responsible for false statements generated by its AI Overview feature — and that AI search is, in the court's own words, "by no means absolutely necessary" for using the internet. The Regional Court of Munich (LG Munich I) issued the temporary injunction on May 28, 2026 (case no. 26 O 869/26), barring Google from repeating false claims about two Munich-based publishers that its AI Overview had linked to scams, subscription traps, and shady business practices. The ruling, first reported by The Decoder, Search Engine Land, Heise Online, Engadget, and The Next Web on June 9–10, 2026, marks a significant escalation in judicial scrutiny of AI search tools globally.
What the Munich Court Actually Decided
The case began when Google's AI Overview conflated information about genuinely disreputable companies with the two Munich-based publisher plaintiffs, generating summaries that falsely associated them with fraudulent business practices. Critically, the court found that none of the false connections appeared in any of the linked source pages — the AI Overview had fabricated the associations itself by mixing and combining information across multiple sources. Before filing suit, the publishers sent Google a cease-and-desist letter through their legal representatives, law firm Lausen Rechtsanwälte, but Google did not respond appropriately, leading to the legal action. The court ordered Google to pay 80% of the legal costs, with each plaintiff publisher bearing 10%.
The legal crux of the ruling is a reclassification of AI Overviews as Google's own speech. German courts have historically given traditional search engines limited liability on the grounds that they merely point to third-party content. The Munich court found that AI Overviews operate categorically differently. According to The Decoder and The Next Web, the court classified Google as a "direct disturber" — not an indirect one — because its AI Overview generates "independent, new, and substantive statements" that constitute Google's own content, not a neutral list of search results.
The court also found that Google cannot invoke host provider protections under the Digital Services Act (DSA) or fall back on the standard notice-and-take-down process applicable to traditional search engines, because the third parties whose websites served as AI source material had not made the false statements in question. As reported by Search Engine Land, the court was explicit: the label "created with AI" does not change the attribution of those statements to Google.
Google's Defenses — and Why the Court Rejected Them
At the hearing, Google argued that users generally understood "that information generated with AI should not be blindly trusted" and could verify claims by checking the linked sources beneath each AI summary. The court rejected this defense outright.
The court drew an analogy to established press law, where publishers are liable for misleading headlines or teasers even if the full article beneath them tells a different or more accurate story. The reasoning is that not every reader follows every link — and the same logic applies to AI summaries that most users will treat as complete, authoritative answers rather than starting points for further research. That concern is supported by data: a Pew Research Center survey from July 2025 found that Google users who saw an AI Overview clicked on a traditional search result link in only 8% of visits, compared to 15% for users who did not receive an AI summary — nearly half the click-through rate.
The court also rejected free-speech protections for AI-generated content. According to The Decoder's translation of the court opinion, the Munich court stated that an AI's output is "not the expression of an acquired conviction of the persons expressing it, but the result of an algorithm," and characterized AI-powered research as "above all an expression of Google's business activities." In the court's view, AI Overviews are a commercial product, not protected speech.
The Accuracy Problem: What the Data Shows
The Munich ruling arrives against a backdrop of growing documented evidence about the reliability — and unreliability — of AI search summaries. In April 2026, AI startup Oumi, commissioned by the New York Times, published an analysis of 4,326 Google searches, testing AI Overviews first with the Gemini 2 model in October 2025 and then again after the upgrade to Gemini 3 in February 2026. The results were mixed.
On the positive side, AI Overviews powered by Gemini 3 were accurate 91% of the time, up from 85% accuracy with Gemini 2. But accuracy alone does not tell the full story. With Gemini 3, 56% of correct AI Overview answers were classified as "ungrounded" — meaning the cited sources did not actually support the information provided. That figure was up from 37% with Gemini 2, suggesting that as the model became better at generating plausible-sounding answers, it became more likely to do so without real evidentiary backing.
The scale implications are stark. Google handles over 5 trillion searches per year. Even assuming AI Overviews are produced for only 50% of those searches, a 9% inaccuracy rate would translate to approximately 225 billion false or misleading summaries per year — roughly 616 million per day, according to TechRepublic's analysis of the Oumi data. Adding to the concern, Google's own internal test data showed that Gemini 3, when running independently of the Google search framework, generated incorrect information in 28% of queries.
Out of 5,380 sources referenced by Google's AI summaries in the Oumi analysis, Facebook and Reddit were the second and fourth most-cited sources respectively — with Facebook appearing as a source in 5% of accurate answers and 7% of inaccurate answers. Google responded to the Oumi findings, with a company spokesperson describing the methodology as having "serious holes." Manos Koukoumidis, CEO of Oumi, countered the implicit reassurance of a 91% accuracy figure with a pointed question: "Even if the answer is correct, how do you know it's correct? How do you verify it?"
How This Fits Into a Larger Legal Pattern
The Munich ruling does not emerge in a vacuum. It builds on an earlier decision by the Frankfurt am Main Regional Court, issued on September 10, 2025 (case ref. 2-06 O 271/25), which involved a group of plastic surgeons whose AI Overview had incorrectly described a medical procedure. In that case, the Frankfurt court dismissed the preliminary injunction after finding the statement was "ultimately not false" when read in complete context — but it also affirmed that Google's liability for AI Overviews is "not per se excluded." The Munich court went further, explicitly prohibiting specific false statements and classifying Google as a direct disturber rather than a passive platform.
According to The Next Web, the Munich court stated that its reasoning could have international reach. Analysts covering the ruling note that the logic applied is not specific to Google's AI Overview — it could in principle extend to any AI answer engine that generates original synthesized responses rather than simply linking to existing content. That would include services like ChatGPT and Perplexity. As of the time of reporting, Google had not commented on the Munich ruling.
Why This Matters Beyond One Court Case
The Munich ruling is significant for several reasons that go beyond the immediate injunction against Google. First, it offers a legal framework — direct authorship rather than passive intermediary — that other courts in Europe and potentially elsewhere could adopt when evaluating AI-generated content. Second, it explicitly rejects the argument that AI systems can disclaim responsibility for their outputs by labeling them as AI-generated. Third, it treats AI-powered search as an optional commercial feature, not a protected utility — a framing that could influence how regulators and legislators approach AI search products going forward.
For users, the practical implication is already visible in the data. The combination of a high rate of ungrounded answers (56% of even correct responses lack genuine source support), a low click-through rate to underlying sources (8% for AI Overview users vs. 15% for non-AI Overview users), and the inherent difficulty of verifying AI-generated claims in real time creates a structural trust problem that a single accuracy upgrade is unlikely to solve. The Munich court's observation that AI search is "by no means absolutely necessary" is a legal finding — but it also reflects a genuine question that users, publishers, and now courts are asking with increasing urgency.
For more tech news, visit our news section.