The legal front: 500+ lawsuits, antitrust, and AI defamation.

The litigation landscape

The publisher-vs-platform litigation cluster is the largest in the history of the open web. 500+ publications have filed or joined lawsuits against AI search platforms. The named cases that matter most for setting precedent:

• The New York Times v. OpenAI & Microsoft — the most watched case; tests whether training and serving from copyrighted content constitutes infringement.
• The New York Times v. Perplexity — separate case, focused on real-time citation rather than training.
• Dow Jones v. Perplexity — joining the wave of suits over verbatim and near-verbatim use of paywalled content.
• EU Commission antitrust investigation of Google — specifically about AI Overviews using publisher content without compensation.
• Wolf River Electric v. Google — the first prominent test of AI-generated defamation against a search engine.

The cases are at different stages, in different jurisdictions, with different legal theories. The cumulative effect, regardless of individual outcomes, is the establishment of legal standards that will shape AI search for the next decade.

The Section 230 question

The most consequential unsettled legal question in AI search is whether Section 230 of the Communications Decency Act — the law that protects search engines and platforms from liability for user-generated content they index — extends to AI-generated outputs.

The American Bar Association’s November 2024 analysis was unambiguous on the uncertainty: “the absence of comprehensive AI regulation clearly defining the contours and applicability of Section 230 immunity in the context of generative AI.” Traditional search engines are intermediaries that display third-party content. AI engines generate new content. The legal distinction matters.

If courts hold that AI-generated content is “speech” by the platform itself rather than indexed third-party content, Section 230 protections likely don’t apply. That would make AI platforms liable for the truthfulness of their outputs in the same way a publisher is liable for its articles. The downstream consequences would reshape the entire economics of AI search.

“If your AI acts as an agent of your business, you likely bear responsibility for what it tells people. — National Law Review”

The Wolf River Electric defamation precedent

The most operationally relevant emerging case for brand-side legal exposure is Wolf River Electric v. Google. The company sued after Google’s AI-generated search results fabricated negative claims about it. The case is one of the first prominent tests of whether AI-generated defamation against a brand meets the legal standard for libel.

The Columbia Law Review’s 2025 analysis, “Redefining Defamation: Establishing Proof of Fault for AI Hallucinations,” framed the central question: traditional defamation law requires the defendant to have acted with knowing falsity or reckless disregard for the truth. How do you apply intent-based standards to a probabilistic system that doesn’t “know” anything?

The Battle v. Microsoft Corp. case is testing a parallel question — an Air Force veteran sued over AI-generated false claims about him. The legal framework for AI defamation is being built case by case in real time. For brands, the implication is twofold: you may have a cause of action when AI lies about you, and your competitors may have one when AI lies about them in your favor. Both directions of risk are operational.

The EU AI Act and labeling requirements

The most concrete regulatory development is the EU AI Act, which began phasing in requirements through 2025 and 2026. The provisions most relevant to GEO operators:

• AI-generated content must be clearly labeled as AI-generated, with disclosure obligations on the platforms generating the content and on entities deploying it.
• Transparency requirements for systems that interact with consumers, including chatbots and AI search interfaces.
• Risk-based classification with stricter requirements for high-risk applications (financial, medical, legal).
• Conformity assessments for AI systems before market entry in regulated categories.

The extraterritorial reach matters: the EU AI Act applies to any AI system used by EU residents, regardless of where the system is operated from. US-based brands optimizing for AI visibility need to factor EU compliance into their content and disclosure strategies.

Individual US states are passing their own AI disclosure laws in parallel, creating a patchwork of state-level requirements. Federal AI legislation has not been enacted, but the FTC has begun applying existing consumer protection authority to AI claims — the May 2025 enforcement actions against companies making deceptive AI claims signal the direction.

FTC enforcement and false-AI-claim risk

The Federal Trade Commission has not waited for new AI legislation. The agency has begun applying its existing consumer protection authority to AI-related deceptive claims. The pattern emerging from enforcement actions:

• $25 million fine against a company that used deceptive AI claims to defraud consumers (recent FTC action).
• Orders barring companies from advertising services dedicated to generating fake consumer reviews or testimonials — a direct strike against the manipulation playbook.
• Joint statement from FTC, EEOC, CFPB, and DOJ clarifying that their existing authority covers AI applications.
• Guidance against falsely claiming AI capabilities, falsely attributing AI-generated content to humans, and using AI to mislead consumers about products or services.

For brand operators: the FTC has signaled it will treat AI-related deception with the same enforcement posture as traditional deception. Brands engaging in synthetic-consensus manipulation, fake-review generation, or hidden-prompt techniques face FTC exposure regardless of whether platforms catch them first.

The right-to-be-forgotten challenge

European data protection law gives individuals a right to be forgotten — the ability to demand the deletion of personal data from systems that hold it. The right has worked reasonably well for traditional search engines, which can de-index URLs.

It does not work for LLMs in any clean technical sense. You cannot simply delete information from a trained model. Information learned during training is encoded in the weights, distributed across billions of parameters, and not extractable in a targeted way. Re-training a model from scratch to remove specific information is computationally prohibitive at the scale of frontier models.

The legal response is still developing. Some platforms offer “output filtering” — preventing the model from producing certain information at inference time even though the underlying knowledge remains in the weights. Whether this satisfies right-to-be-forgotten requirements is being litigated. Brands and individuals seeking to remove inaccurate or outdated information from AI engines face a fundamental technical barrier the legal framework hasn’t yet caught up to.

The negotiation alternative

Some publishers have chosen to settle rather than litigate. News Corp signed a $50M-per-year licensing agreement with OpenAI. The Associated Press has a similar deal. Other major outlets have followed.

The math behind the choice: a $50M annual deal is meaningful for News Corp’s revenue base and creates a multi-year predictable revenue stream during a period of structural traffic decline. Litigation would cost less but might yield nothing — and would not stop the traffic loss while the case wound through the courts.

The economics of negotiation favor only the largest publishers. Mid-tier publishers are not getting offered comparable deals. Local news, trade publications, and most B2B media are litigating or absorbing the loss with no realistic path to a licensing solution. The result is a two-tier publisher ecosystem: a small number of tier-1 outlets with AI revenue streams, and everyone else competing for what remains of the open-web traffic model.

What every brand’s legal team should brief on

The operational read for in-house legal teams supporting marketing and brand functions:

1. Vendor due diligence on GEO partners. Any vendor whose tactics include synthetic personas, fake reviews, astroturfing, or hidden prompt injection is exposing your brand to FTC and (in the EU) AI Act enforcement.
2. Content disclosure obligations. AI-generated content used in customer communications may require labeling under EU and emerging US state law. Your content production process should track AI involvement at a granular level.
3. Defamation monitoring and response. Establish a process for identifying AI hallucinations about your brand and documenting them for potential legal action. The Wolf River and Battle cases are establishing precedent that may give you remedies.
4. Section 230 watch. The legal status of AI-generated content is being established this year and next. Your brand’s exposure to AI-generated misinformation may shift dramatically depending on how courts rule.
5. Right-to-be-forgotten requests. If your brand or executives need to remove inaccurate information from AI engines, the technical and legal pathways are limited. Plan for slow timelines and partial remedies.
6. Licensing deal evaluation. If your business depends on content that AI platforms are training on, the licensing-vs-litigation question may eventually apply to you. Track the evolving deal terms in publishing as a forward indicator.

Every statistic in this piece follows Visigent's research standards: named sources, dated figures, corrections on request to help@visigent.com.