AI visibility work has acquired a content-rights layer that cannot be treated as a separate legal-office problem. Editorial Perfil has sued OpenAI and Microsoft in Argentina over alleged unauthorised use of its journalism, while News Corp is suing browser company Brave over alleged scraping and monetisation of copyrighted reporting. These are allegations, not findings, but their practical significance for GEO is immediate: the path from publisher page to AI answer is becoming commercially contested.
The useful response is not to assume that every use of public web material is prohibited, nor to assume that a page being reachable makes it a safe input for every AI workflow. The supplied reporting does not settle those questions. It does show publishers challenging scraping, training, answer generation and competing uses of their work at the same time as some publishers sign paid agreements with AI companies. That combination makes provenance, permissions and partnership choices part of AI-visibility strategy.
Two cases, different routes from journalism to AI products
Editorial Perfil said it filed suit in Argentina’s Federal Civil and Commercial Courts against OpenAI, Microsoft and Microsoft’s local subsidiary. The publisher alleges that its newsroom content was used without authorisation or payment to develop and train commercial AI systems, and alleges copyright infringement and unfair competition. Perfil says this included articles, investigations and interviews, including material behind its paywall, that were tracked, downloaded and stored through large-scale scraping. Buenos Aires Times
Perfil also alleges that products including Bing, Copilot, Microsoft 365 Copilot, GitHub Copilot and Azure AI answer questions directly using journalistic content, reducing the need for users to visit news sites. The publisher connects that alleged behaviour to traffic, advertising revenue and the financial sustainability of journalism. Neither OpenAI nor Microsoft had publicly responded to the suit at the time the report was written. Buenos Aires Times
The News Corp case concerns a different alleged intermediary. News Corp sued Brave AI, alleging that the browser company scraped copyrighted journalism without authorisation and sold it to AI companies. According to the report, News Corp claims Brave disguised web crawlers to bypass publisher restrictions and supplied “verbatim or near verbatim” summaries of News Corp content before March 2025. IBTimes
Brave disputes the broader framing. It had separately sued News Corp after receiving a cease-and-desist letter, arguing that indexing website content is essential to operating a search engine and accusing News Corp of trying to push it from the market for the benefit of larger technology companies. IBTimes
For GEO practitioners, the distinction matters. A dispute about model training is not identical to one about crawling, retrieval, indexing, summarisation or a browser-generated answer. But these cases show that publishers may regard several links in that chain as a single commercial problem when the result substitutes for a visit, competes with a licensing business, or monetises their work.
The licensing market is not a clean answer
The same reporting shows the market splitting between litigation and licensing rather than resolving into one standard practice. Perfil says OpenAI has signed licensing agreements with large international media groups, including Spain’s Prisa Media, while alleging that Argentine and Latin American outlets have not been compensated for content used. OpenAI has publicly said journalism is important to its products and that it has agreements with organisations including the Associated Press, Axel Springer, the Financial Times, Le Monde, Prisa Media, Vox Media, Condé Nast and Reuters. Buenos Aires Times
IBTimes also reports that News Corp entered a multi-year Meta agreement worth up to $50 million annually, allowing Meta products to use archive content and retrieve new information from News Corp publications. It lists other recent agreements involving Microsoft and Australia’s Nine Entertainment; OpenAI and The Washington Post, Axios, Time and Hearst; and Amazon with Reach, Condé Nast and Hearst. IBTimes
That does not mean a licensing deal guarantees prominence in an AI answer, and the sources provide no evidence that it does. Nor do they establish that an unlicensed publisher will be excluded from AI systems. What the evidence supports is narrower: publisher relationships can now be both a distribution arrangement and a disputed boundary around AI use.
Why GEO teams should change their operating model
GEO often treats sources as signals: authoritative pages to earn mentions from, material to analyse for answer patterns, or content to make accessible to AI retrieval. The current publisher conflict adds a question before all three: what rights and access conditions attach to the source, and what use is the team or its vendors actually making of it?
That should change the internal brief. A proposal that says “use leading publisher coverage to improve AI visibility” is incomplete if it does not distinguish between earning independent editorial coverage, quoting short excerpts in an internal research document, ingesting content into a retrieval system, generating summaries, or reselling outputs. These uses are not equivalent in the allegations now being made by publishers.
- Map provenance. Keep a record of which publisher materials informed a GEO corpus, prompt library, knowledge base or retrieval workflow.
- Separate discovery from reuse. Finding a page in search or an AI answer does not answer whether its full text can be stored, transformed or used in a product.
- Record restrictions and permissions. The Brave allegations specifically concern bypassing publisher restrictions, making access controls a material operational signal rather than an implementation inconvenience. IBTimes
- Define partnership claims precisely. Do not present a publisher licence as proof of answer inclusion, referral traffic or AI visibility unless the agreement and evidence actually support that claim.
This is particularly important when agencies and in-house teams use third-party AI platforms. A team may not operate the crawler or train the model, but it can still choose sources, upload material, commission summaries and represent the resulting workflow to a client. The sources do not establish liability for any of those actions. They do establish that publishers are contesting the economic use of journalism across the AI stack.
The litigation volume makes this a durable planning issue
Originality.ai’s tracker listed 50 cases against OpenAI and ChatGPT as of July 2026. Its analysis says 56% involved copyright or training data, and that creators, publishers and data owners were the largest plaintiff group, with 28 cases. The tracker is not a court record or legal advice, and it covers OpenAI-related cases rather than all AI disputes. Still, it is useful evidence that copyright and training-data claims are not isolated to one publisher or one jurisdiction. Originality.ai
The same tracker says 12 cases were centralised for coordinated pretrial handling in the OpenAI copyright multidistrict litigation in April 2025, while explaining that centralisation does not merge claims into a single case. Originality.ai That procedural detail matters because it undercuts any temptation to read one complaint as a settled rule for the entire market. The legal outcomes remain unsettled in the supplied sources.
The strategic conclusion is therefore disciplined rather than dramatic. GEO cannot promise that useful publisher content will remain freely available for AI use, that licensed sources will win answer visibility, or that litigation will produce one global rule. It can recognise that sourcing decisions now carry partnership and rights risk. Teams that document source provenance, respect stated restrictions and avoid overstating what a relationship with a publisher delivers will be better positioned than teams still treating the web as an undifferentiated training and retrieval layer.