The relationship between publishers and artificial intelligence companies is shifting from confrontation to commerce. As AI systems need reliable, current, and large-scale text to train models and power live answers, publishers are increasingly choosing to license their journalism rather than rely only on lawsuits. In practice, this means publishers sell data to AI content generators through negotiated agreements that turn archives, reporting, and brand trust into monetizable assets.
This trend accelerated in 2025 and 2026 as major technology companies sought better access to verified reporting, while news organizations looked for fresh revenue to offset declining traffic and the growing threat of AI-generated summaries. The result is a fast-evolving market in which licensing, usage rights, crawler access, and payment models are becoming central to the future of digital publishing.
A market rapidly moving from scraping to licensing
For years, many publishers accused technology companies of ingesting content without permission to improve AI systems. That dynamic created tension across the media industry, especially as generative AI tools began summarizing articles, answering news questions, and reducing the need for users to click through to original sources. In response, licensing has emerged as a more structured alternative to unapproved data collection.
Reuters reported in 2025 that Meta had struck multiple commercial licensing agreements with news publishers to give Meta AI access to current, verified reporting. The move was presented as part of a wider push to bring real-time news into AI products. This is important because AI firms increasingly need not just old web data, but timely and trustworthy journalism that can support fresh answers.
At the same time, publishers see these deals as a way to reclaim value from their work. Rather than allowing AI systems to benefit from reporting without compensation, media groups are seeking direct payments for access, usage, and visibility. That is one reason the debate has expanded from simple training-data questions to broader discussions about pay-to-use and pay-to-crawl models.
Meta's deals signaled a major shift
One of the clearest signals of this new phase came from Meta. In late 2025, the company reportedly signed multiple AI data agreements with news publishers including USA Today, People Inc., CNN, Fox News, The Daily Caller, Washington Examiner, and Le Monde. The report suggested that Meta was returning to compensating news organizations after previously stepping back from such arrangements.
That change matters because Meta had long been seen as cautious about direct publisher payments in the news space. By moving back into deals, the company effectively acknowledged that quality journalism remains a valuable input for AI systems. The new agreements also suggest that major AI platforms now recognize a practical need for licensed, professionally produced content rather than relying solely on open-web collection.
News Corp's own agreement with Meta underscored the scale of the opportunity. According to reporting, the deal could be worth up to US$50 million per year. News Corp’s CEO described news organizations as a critical “input” for AI, while also making clear that the company would pursue legal action if its content were taken illegally. That balance between partnership and enforcement is becoming a defining strategy across the industry.
The archive is now a strategic asset
Publishers are not only selling access to current reporting. They are also monetizing decades of archived journalism, which is especially valuable for training large language models. Historical reporting provides volume, language patterns, factual context, and editorial consistency, making archives a powerful asset in AI negotiations.
The Associated Press illustrated this clearly. AP said its 2023 licensing deal with OpenAI gave the AI company access to AP news stories going back to 1985. AP presented the agreement as a way to license its archive at a time when other tech companies were absorbing enormous quantities of written material to improve AI systems without always offering publishers a clear return.
That strategy expanded into current-news licensing as well. In January 2025, AP reported that Google’s Gemini would deliver up-to-date news from The Associated Press in Google’s first such deal with a news publisher. Together, these arrangements show how publishers can commercialize both the depth of their archives and the immediacy of their live reporting.
New marketplaces are making AI licensing easier
Beyond one-on-one contracts, a broader infrastructure for content licensing is now emerging. In February 2026, Microsoft publicly launched its Publisher Content Marketplace, presenting it as both a revenue stream for publishers and a mechanism for AI companies to license quality content more efficiently. Microsoft said it had already worked with publishers such as Business Insider, Condé Nast, Hearst Magazines, People, The Associated Press, USA Today, and Vox Media on licensing, pricing, governance, analytics, and onboarding.
This development suggests that AI licensing is becoming a formal market rather than a series of isolated deals. Standardized tools for pricing, governance, and measurement could help publishers of different sizes participate, not just the largest global brands. It also gives AI developers a more organized way to obtain legitimate access to trusted material.
Dow Jones has been building in the same direction. Axios reported in February 2025 that Dow Jones had created an AI marketplace for publishers through Factiva, with nearly 5,000 publishing partners. That figure had risen sharply from nearly 4,000 in November and roughly 2,000 six months before launch, indicating strong publisher interest in collective licensing systems.
How payment models are evolving
At first, many publisher-AI agreements were discussed in terms of flat fees for model training. But the market is becoming more sophisticated. The Information reported that publishers licensing content to AI firms were increasingly talking about usage-based payment models instead of one-time training fees, reflecting concern that AI systems continue generating value long after initial ingestion.
This shift makes sense in a world where AI tools can repeatedly summarize, quote, or synthesize publisher content in chatbots and search interfaces. If a model uses journalism continuously to answer user questions, publishers argue they should be compensated continuously as well. That is why negotiations increasingly include recurring fees, output-related terms, and crawler-based compensation.
Industry reporting in 2025 and 2026 described a broader move toward pay-to-use and pay-to-crawl discussions. Cloudflare data cited by The Information showed traffic from one OpenAI crawler increased 305% between May 2024 and May 2025. For publishers, that kind of growth reinforces the idea that access itself has value and should be governed commercially rather than left unmanaged.
Licensing is rising, but litigation still matters
Even as publishers sell data to AI content generators, legal risk remains a major factor in negotiations. Many media companies have concluded that licensing can produce faster and more predictable revenue than courtroom battles, but that does not mean lawsuits are disappearing. Instead, litigation often serves as leverage in dealmaking.
The Guardian described News Corp’s approach as “woo or sue,” meaning the company is willing to partner with AI firms while also taking legal action when content is used without authorization. In March 2026, News Corp’s CEO repeated that formula, using it to summarize the balance between welcoming commercial agreements and defending copyright aggressively. This captures the mood of much of the industry.
A key legal milestone came in February 2025, when Thomson Reuters won an early U.S. court victory in an AI copyright case involving unauthorized use of its Westlaw material. That outcome showed publishers they may have viable claims, but it also underscored why many prefer licensing over relying solely on litigation. A contract can create revenue now, while a lawsuit may take years.
An industry split is turning into a hybrid strategy
Several major publishers and media groups have already chosen licensing over pure litigation, including the Financial Times, Axel Springer, the Guardian, Schibsted, and others. This does not necessarily mean they are comfortable with every aspect of AI development. Rather, it reflects a practical recognition that AI companies need quality journalism, and that publishers can use that need to negotiate compensation.
AP has expressed the core tradeoff clearly. Publishers are at a disadvantage when technology companies integrate AI-generated summaries into search and chat tools, because those summaries can divert audience attention and reduce traffic to original sites. Yet licensing deals can bring badly needed revenue and may improve information quality if AI systems are built on verified reporting instead of lower-quality material.
That hybrid strategy is likely to define the next phase of the market. Publishers are no longer choosing strictly between selling access and fighting in court. Many are doing both: monetizing approved use while pushing back against unauthorized use. In that sense, licensing is not replacing copyright enforcement; it is being added to it.
Standards and collective pressure are becoming essential
As more deals are signed, publishers are also trying to shape the rules of the market. In February 2026, a coalition called Spur was launched by major news brands including the Guardian, BBC, Financial Times, Sky News, and Telegraph Media Group. The group said it wants AI developers to access journalism through legitimate, responsible, and convenient licensing frameworks while ensuring publishers are paid fairly.
This kind of collective effort shows that individual contracts alone may not solve the deeper structural issues. Publishers still worry about transparency, attribution, fair payment, crawler control, and the possibility that AI systems will reduce direct relationships with readers. Shared standards could help define what responsible AI access looks like across the industry.
The same concerns are visible outside Europe and North America. In February 2026, Nine Entertainment said it had struck two licensing deals with key domestic corporates for use of its content to train in-house large language models. At the same time, it urged the Australian government to prioritize technology compensation rules in the news media bargaining code process, showing that commercial agreements and regulatory pressure are advancing together.
The rise of these agreements shows that publishers are not simply passive victims of the AI boom. They are increasingly treating journalism, archives, and metadata as premium assets that can be licensed in structured ways. As platforms race to improve their models with current and credible information, publishers have a stronger bargaining position than they did in the early wave of AI development.
Still, the long-term outcome remains uncertain. If publishers sell data to AI content generators without securing strong usage rights, recurring payments, and clear standards, they could help build products that further weaken their own traffic and influence. But if they negotiate effectively, licensing could become one of the most important new business lines in modern media, turning the value of trusted reporting into a sustainable source of revenue.
