On 28 May 2026 the Munich Regional Court I (Landgericht München I) ordered Google, in preliminary injunction proceedings, to stop stating in its AI Overviews that a Munich publishing group was linked to scams and subscription traps. The reason: the AI-generated summary was Google’s own content, not merely a list of third-party results. For brand owners, the case shows that AI Overviews liability in Germany can be enforced like any other false statement of fact.
Update (October 2026): according to the Bavarian official case-law portal, Google appealed to the Higher Regional Court of Munich (case 18 U 1744/26) and the parties then ended the dispute by settlement, so the judgment no longer has effect. The court’s reasoning remains the most detailed published analysis of the issue so far. Bayern.Recht, LG München I, 26 O 869/26.
Key takeaways
- The court treated Google’s AI Overview as Google’s own statement, not as a search result, so Google was liable as a direct infringer.
- The hosting privilege in Article 6 of the EU Digital Services Act (DSA) did not protect Google, and Article 6(4) DSA expressly allows national courts to order a provider to stop an infringement.
- Decisive facts: the summary said things that none of the linked sources said, and Google did not check it after being notified.
- The judgment was a first-instance decision in summary proceedings, not a ruling from a higher court.
What did the Munich court decide?
The applicants were a Munich publishing house with twelve imprints and one of its subsidiaries. In January 2026, a Google search for the company name plus “Betrugsmasche” (scam), a term suggested by Google’s own autocomplete, produced an AI Overview presenting the publisher as known for dubious practices and subscription traps, and linking it with unrelated companies. The publisher sent a cease-and-desist letter on 2 February 2026; Google referred it to its online form.
In its judgment of 28 May 2026 (case 26 O 869/26), the 26th Civil Chamber granted most of the injunction under sections 1004 and 823(1) of the German Civil Code, read with the constitutional protection of a company’s reputation (the “corporate personality right”). It rejected some of the requested wording and split the costs 80% to Google and 20% to the applicants.
Why was Google liable for AI Overviews in Germany?
German case law treats a search engine as an indirect contributor to third-party content: it must act only once it is told about an obvious infringement (Federal Court of Justice, VI ZR 489/16, 2018). The Munich court refused to apply that rule to AI Overviews:
| Issue | Classic search results | AI Overview, as the court saw it |
|---|---|---|
| Who is speaking? | Third-party websites; the engine only links | Google, which summarises “in its own words” and with its own structure |
| Content | What the sources say | New statements; here, links between the applicants and other firms that no linked source made |
| Duty to check | Only after notice of an obvious infringement | Checking is feasible by comparing the summary with its sources, and Google did not do so even after notice |
| Essential for the internet? | Yes, which justifies limited duties | No; search works without the AI feature |
The court also rejected the argument that users can click through and verify. Like a newspaper headline read in passing, the AI Overview reads as a complete answer and gives no sign that it may be unreliable. Neither the GDPR (the applicants were companies) nor the EU AI Act, which only adds an administrative complaint route, displaced these civil claims. The chamber added a practical point, highlighted by the IPKat on 28 June 2026: if Google were not liable, the affected company would have no one to sue, because the websites behind the links never made the statements.
What this means for your business
The order was limited to Germany and to one set of facts, but the logic travels: if an AI system writes a new sentence about your company, the operator may be treated as its author. In practice:
- Monitor AI answers. Add AI Overviews and chatbots to your brand watch, using your company name, your marks and risk terms such as “scam” or “fraud”.
- Capture evidence properly. Dated screenshots, the exact query and full result pages carried weight with the Munich court.
- Notify through every channel: the provider’s own form and a formal letter. Notice was part of what made Google’s inaction decisive.
- Find the root cause. Here the AI mixed in information about other firms. A clear, registered brand and accurate public information reduce that risk.
- Move quickly. Preliminary injunctions in Germany depend on urgency, and delay can cost you the remedy.
Our team for AI and digital asset protection can add AI-generated answers to your brand monitoring and prepare the evidence file.
Where companies get this wrong
- Treating AI errors as unfixable. The Munich case shows they can be the subject of a court order, at least in Germany.
- Relying on a single complaint form. Google argued that the notice was insufficient; a reasoned letter that explains why each statement is false is harder to dismiss.
- Failing to prove falsity. Google pointed to negative online reviews. You must show which statements are untrue, not just that they are unpleasant.
- Ignoring confusion with other businesses. Information about other firms fed the error, which is where trademark registration and watch services help.
- Waiting for the system to correct itself. The text may change, but the harm is done and urgency may be lost.
Frequently asked questions
Is Google always liable for its AI Overviews in Germany?
No. This was a first-instance decision in summary proceedings by the Munich Regional Court I, based on statements that the linked sources did not contain and on Google’s failure to act after notice. Other German courts may take a different view, and no higher court had ruled on the question when this judgment was issued.
Can a company outside Germany use this decision?
Possibly, if its centre of interests is in Germany or the false statement causes harm there. The court based its jurisdiction on Article 7(2) of the Brussels I bis Regulation and applied German law because the injury occurred in Germany. In other countries, national defamation and unfair competition rules will decide the outcome.
What evidence do I need against an AI-generated statement?
Dated screenshots of the full result page, the exact search query, proof of what the linked sources actually say, and evidence that the statements are false. Keep copies of your notice to the provider and any reply. The Munich court relied on this type of material to accept the applicants’ account.
Can IP Global Guard handle a complaint about a false AI answer?
Yes. We monitor AI answers alongside your brand watch, prepare the evidence and the notice to the provider, and coordinate qualified local counsel where court action is needed, in Germany or elsewhere in Europe, Latin America and Africa, from a single point of contact.
How IP Global Guard protects your reputation in AI answers
IP Global Guard, the IP services line of META Channel Corporation Limited, combines brand monitoring, trademark portfolio work and enforcement across more than 25 jurisdictions in Europe, Latin America and Africa, with one strategy and one billing relationship. Within the same group, our regulatory colleagues cover the EU AI Act and the GDPR.
If an AI system is saying something false about your company or your brand, send us the screenshots, the search terms and the markets where it appears. We will assess what can be done and in which country. Contact our team.
This article is general information, not legal advice, and reflects the position on the date of publication.
