Germany's media regulator ruled on July 14, 2026 that Google's AI Overviews and Perplexity are publisher-created content rather than neutral search results, removing the standard platform liability shield for AI-generated answers. The decision reframes a core assumption of the AI search era: the engine is now responsible for what it says about you.
For most of the search era, the legal deal was simple. Search engines pointed at other people's content and were generally shielded from liability for what that content said. AI search broke that arrangement quietly, because an AI Overview does not point at a source. It writes a new answer. Germany is the first authority anywhere to say, formally, that this difference matters in law.
What Germany Actually Ruled
The direct answer: Germany's Commission for Licensing and Supervision (ZAK), which coordinates the country's state media authorities, issued administrative rulings that Google AI Overviews and Perplexity are subject to German media law. As The Decoder reported, regulators concluded these services operate as content publishers, not neutral conduits, which means they cannot claim the standard intermediary liability exemption that shields ordinary search results.
The rulings followed investigations by the Hamburg/Schleswig-Holstein and Berlin-Brandenburg media authorities, and they attach obligations around accuracy, transparency, and media plurality. According to MediaPost, the regulator's argument is that AI-generated summaries and chatbot outputs constitute provider-created content, which places them outside certain protections available to intermediaries under the EU's Digital Services Act. Both companies have an appeal window, and the case would then proceed through German administrative courts.
This administrative ruling did not appear in a vacuum. Weeks earlier, the Regional Court of Munich granted a temporary injunction (case no. 26 O 869/26) barring Google from spreading false claims about two Munich publishers through AI Overviews. As documented by the Library of Congress Global Legal Monitor and Malwarebytes, the AI feature had falsely tied the two companies to scams and subscription traps, mixing up information and inventing connections that appeared in none of the linked sources. Google has said it will appeal.
Why "Publisher, Not Conduit" Is the Phrase That Matters
The specific companies named are less important than the legal reasoning. The Munich court classified Google as a direct infringer because the AI Overview is its own content, generating what the court called "independent, new, and substantive statements" by combining material from third-party sites. Crucially, the court held that a disclaimer telling users to double-check AI results is not enough to escape responsibility for what those results assert.
Strip away the German legal specifics and the principle is portable: an AI system that writes an original answer is speaking, and the entity that operates it may own that speech. That is a very different liability posture than "we just listed links." For any business, this cuts two ways at once, and both deserve attention.
The First Edge: Your Brand in Someone Else's AI
The immediate business risk is reputational. AI Overviews, ChatGPT, and Perplexity now sit between your company and a large share of the people researching it, and they summarize you in their own words. When that summary is wrong, the damage is real, and until recently the practical recourse was thin.
The Munich case is the template for how that changes. Two publishers were described as connected to scams by an AI feature drawing on sources that said no such thing, and a court ordered the output corrected. This is the accountability counterpart to the visibility problem we covered in how AI Overviews are reshaping the way people find your business. Visibility asks whether AI engines mention you. Accountability asks whether what they say is true, and who answers when it is not.
The practical response is not to wait for a lawsuit. Businesses that manage this well are already monitoring how the major AI engines describe them, capturing dated screenshots when an answer is false or defamatory, and using each platform's correction and feedback channels. Because these engines synthesize from public sources, the durable fix is upstream: keep accurate, consistent, well-structured information across your own site and the authoritative directories these systems trust, so the model has correct material to summarize in the first place.
The Second Edge: Your Own AI Is Speaking Too
The more strategic implication points inward. If regulators and courts increasingly treat AI-generated statements as the operator's own speech, that logic does not stop at Google. It eventually reaches the customer-facing assistant on your own website.
A support bot that quotes the wrong refund policy, an AI sales agent that promises a discount that does not exist, or a knowledge assistant that fabricates a product specification is producing statements your business may own, in the same way Google was found to own its Overview. "The AI said it, not us" is precisely the defense the Munich court rejected. As the "publisher, not conduit" principle spreads, the accuracy of a deployed AI system stops being a quality-of-experience nicety and becomes a liability surface.
This is why serious teams are moving accuracy and source grounding out of the prompt-tuning afterthought category and into production systems engineered to answer only from verified company data, with retrieval constraints, confidence thresholds, and clear escalation to a human when the system is unsure. Our take: the era of shipping a customer-facing chatbot that free-associates over a base model's training data is ending, not because it is embarrassing, but because it is becoming a documented risk.
What This Means for Your Business
The through line connecting both edges is that AI-generated content is losing its "nobody is responsible" status. That reframing should shape how you treat AI in your stack today.
- Externally, treat AI search as a channel you must manage, not just appear in. Monitor your brand's AI-generated descriptions the way you would monitor reviews, and act on falsehoods with documentation rather than resignation.
- Internally, treat your own AI's statements as your statements. Assume that anything your deployed systems assert to a customer could one day be attributed to your company, and design accordingly.
- Build the record now. Log what your AI systems say and on what basis, because the ability to show grounded, source-backed behavior is the difference between a defensible system and an indefensible one. This discipline is the same backbone we described in our practical AI governance framework.
A necessary caveat: the German rulings are jurisdiction-specific, they are under appeal, and none of this is legal advice. The strategic signal, however, travels well beyond Germany. Regulators across the broader AI compliance landscape are circling the same question of who answers for what an AI says, and the direction is consistent.
Common Mistakes to Avoid
The first mistake is dismissing this as a German edge case. The specific ruling is local, but the legal reasoning, that an AI answer is original authored speech, is exactly the argument that tends to migrate across jurisdictions. The second mistake is treating brand accuracy in AI search as a marketing problem alone; it is now also a legal and reputational one. The third, and most expensive, is assuming disclaimers protect your own AI. A court has now said, in plain terms, that "please verify this" does not transfer responsibility for a false statement back to the reader.
Key Takeaways
- Germany's ZAK ruled on July 14, 2026 that Google AI Overviews and Perplexity are publisher-created content, removing the standard platform liability shield for their AI-generated answers.
- A Munich court separately ordered Google to stop spreading false AI Overview claims about two publishers, rejecting the idea that a "verify this" disclaimer transfers responsibility.
- The reusable principle is that an AI system writing an original answer is speaking, and its operator may own that speech.
- Externally, this gives businesses a stronger footing to demand corrections of false AI-generated claims about them.
- Internally, it signals that a company may increasingly be held accountable for what its own customer-facing AI states, raising the bar for accuracy and grounding.
The businesses that move early on AI accountability will have a meaningful advantage as the rules settle. If you want to be one of them, let's start with a conversation.