German Court Rules Google Liable for False AI Overviews
AI Overviews are Classified as Original Content, Not Search Results
The Regional Court of Munich has ruled that Google is directly liable for the accuracy of its AI-generated search overviews. In case no. 26 O 869/26, the court issued a temporary injunction barring Google from spreading false claims about two Munich-based publishers. The core of the ruling is the distinction between a traditional search engine and an AI-driven summary: because the AI rewrites and structures information in its own words, the resulting output is Google's own content, not a curated list of external links.
In this specific case, Google's AI Overviews falsely linked the plaintiffs to scams and subscription traps. The court found that the AI mixed up information from other companies and created connections that did not exist in any of the linked sources. Consequently, the court classified Google as a direct infringer.
Search Engine Liability Protections Do Not Apply to AI
Traditional search engine operators in Germany have historically benefited from limited liability under Federal Court of Justice (BGH) rulings, which viewed them as indirect infringers who merely made third-party content findable. The Munich court explicitly rejected the application of these protections to AI Overviews for several reasons:
- Independent Statement Generation: Unlike traditional search, which points to outside websites, AI Overviews generate "independent, new, and substantive statements" by evaluating and combining third-party content.
- Exclusive Control: The court noted that only Google has influence over the algorithms and the offering, meaning only Google can verify the accuracy of the AI's statements by comparing them to the underlying sources.
- Lack of Necessity: The court determined that AI Overviews are an "extra feature" and not absolutely necessary for internet usage, as traditional search results already provide the required utility.
Because of these factors, Google cannot invoke host provider protections under the Digital Services Act or rely on standard notice-and-take-down processes.
Rejection of the "User Verification" Defense
Google argued that users are aware that AI-generated information should not be blindly trusted and can verify claims by clicking the provided source links. The court rejected this defense, stating that the ability to disprove a statement through further research does not exempt a provider from liability for making that statement.
The court highlighted that AI Overviews are "understandable on its own" and present self-contained statements. This reasoning is supported by data indicating that only 1% of users click a source link directly from an AI Overview. The court compared this to press law, where publishers are held liable for "teasers" that are understandable on their own, regardless of whether the reader accesses the full article.
Reduced Free Speech Protections for Algorithmic Output
The ruling further limits the legal protections typically afforded to free speech. The court determined that an AI's "opinion" is not the expression of a conviction held by a person, but rather the result of an algorithm. Therefore, providing AI-powered research is viewed primarily as a business activity rather than an expression of belief or opinion.
When weighing the privacy and reputation rights of the plaintiffs against Google's business interests, the court ruled that the plaintiffs' rights took precedence, especially since the AI's claims were based on untrue facts.
Broader Implications for AI Providers
This ruling creates a significant legal precedent that could extend beyond Google to other AI providers such as OpenAI, Anthropic, and Perplexity.
Scale of Error
An analysis by AI startup Oumi for the New York Times found that while Gemini 3 (powering AI Overviews) is correct 91% of the time, the remaining 9% represents millions of incorrect answers per hour given Google's scale. Furthermore, 56% of correct answers could not be backed up by the linked sources, illustrating the "black box" nature of the AI's reasoning that the Munich court addressed.
Industry Perspectives
Community discussion on Hacker News reflects a divide between those who view this as a necessary correction and those who fear it may stifle AI deployment in Europe.
"The true mark of AGI is when a company accepts liability and doesn’t bury ‘for entertainment purposes only’ deep in their TOS."
Conversely, some observers worry that the inability to guarantee 100% accuracy across millions of calls could lead to a "soft ban" on non-deterministic software in Germany or the EU:
"If companies can be held liable... for the output of non-deterministic software, isn't this just a soft ban on the deployment of non-deterministic software?"
Google was ordered to cover 80% of the legal costs in this case, and the ruling may have international reach if the reasoning is adopted by other jurisdictions.