German Court Rules Google Liable for Defamatory AI Overviews in First-of-Its-Kind Decision

A German court has ruled that Google can be held directly liable when its AI Overviews produce false and damaging statements, marking the first decision anywhere to classify an AI search summary as the platform’s own commercial speech rather than a neutral container for third-party content. The preliminary injunction from the Regional Court of Hamburg (Landgericht Hamburg) found that the defamatory claims did not appear in any source the AI summarized, meaning the model invented the smear on its own. Google says it is "carefully reviewing this decision, which is not yet final," and an appeal is widely expected.

What Did the Hamburg Court Actually Decide?

The dispute centered on an AI Overview that generated defamatory claims about a publisher. Crucially, the court found that the false statements did not appear in any of the web pages the AI had summarized. The model drew what the court described as "independent, new, and substantive statements" from its own misreading of the underlying links, and that distinction is the heart of the ruling.

Why are AI summaries treated as Google’s own speech?

The court characterized the AI Overview’s output as "primarily an expression of the defendant’s commercial activity." That framing strips away the intermediary shield that search engines have long relied on. A list of blue links points elsewhere. A generated paragraph is an original authored statement, and the author here is Google. Because only Google can change the algorithm that produced the false claim, only Google can be held responsible when that claim causes harm.

Is AI search considered essential to using Google?

The court rejected the idea that AI Overviews are an essential function of search. It described the feature as "an additional function, one without which the use of the search engine would still be (and is) possible, and without which users are perfectly capable of finding results amidst the flood of data." Because the tool sits on top of an already-functional search engine, the platform cannot claim immunity on the grounds that curating information at scale is inherently difficult.

Do disclaimers shield Google from liability?

The court dismissed the "do not blindly trust us" defense, noting that the feature’s usefulness "would be significantly diminished if the AI overview were generally regarded as unreliable and if every single displayed link required independent verification." A company cannot market a feature as trustworthy and simultaneously argue that nobody should trust it. If the business model depends on users relying on the output, a disclaimer does not erase legal responsibility for that output.

What Are the Ruling’s Key Takeaways?

• Generated text is authored speech. When an AI writes a new sentence about a person, company, or event, the platform that operates it is treated as the speaker.
• Liability follows control. The party that controls the model that produced the statement is the party that answers for it. There is no original source to redirect blame toward when the AI fabricated the claim.
• "It is just a tool" is not a shield. Optional, value-added features do not inherit the legal protections that core indexing and linking once enjoyed.
• Trust cuts both ways. Selling a product on the promise of reliability undercuts the defense that users should have assumed it was unreliable.

What Happens Next?

Google has said it is "carefully reviewing this decision, which is not yet final," and an appeal is widely expected. Even if the injunction is overturned, the precedent will not disappear quietly. It is the first time a court has tested the theory that AI-generated speech equals platform liability, and the framework is now part of the legal record that lawyers elsewhere can cite.

The reasoning is especially relevant inside the European Union, where the AI Act already imposes a strict liability framework on high-risk and consumer-facing AI systems. Other EU courts working under that regime may find the Hamburg logic persuasive. The exposure is not limited to search either. Any product that publishes generated text about real entities, such as automated product descriptions, review summaries, comment digests, or suggested replies, faces the same theory of harm. Expect platforms to tighten internal review and, in some cases, to pull back on fully automated generation until the liability picture clarifies.

The Bigger Picture for Businesses and Public Figures

For years, the dominant legal posture treated search engines as passive conduits. They pointed at content; they did not create it. Generative AI collapses that distinction. A model that composes an original paragraph is not pointing at anything. It is publishing, and the Hamburg court has now said plainly that publishing carries consequences regardless of who or what does the writing.

That has practical weight for anyone whose reputation can be summarized by a machine, which is to say nearly every business and public figure. A fabricated claim inside an AI Overview is not a quirky model error to be shrugged off. It is a published statement with a legally identifiable author. The defensive posture that follows is unglamorous but clear: monitor how AI search features describe you, document any false or defamatory output the moment it appears, notify the platform, and bring in counsel when real harm occurs. Speed matters, because a generated smear can propagate through downstream systems before anyone notices it was never true to begin with.

One court has now told the largest search engine on earth that AI-generated speech is real speech, and that real speech comes with real consequences. Whether or not this specific injunction survives appeal, that idea is unlikely to be unlearned.

FAQ

What did the German court decide about Google’s AI Overviews?

The Regional Court of Hamburg ruled that Google can be held directly liable when its AI Overviews produce false and defamatory statements, classifying the generated summaries as Google’s own commercial speech rather than neutral third-party content.

Were the defamatory claims found in the original web pages?

No. The court found that the false statements did not appear in any of the web pages the AI had summarized. The model drew "independent, new, and substantive statements" from its own misreading of the underlying links, meaning the smear was invented by the AI itself.

How has Google responded to the ruling?

Google has said it is "carefully reviewing this decision, which is not yet final," and an appeal is widely expected. The preliminary injunction is jurisdiction-specific, but the reasoning could influence other EU courts operating under the AI Act’s strict liability framework.