The New York Times is taking a selective approach to licensing its journalism to artificial intelligence companies, and its criteria reveal a larger challenge facing publishers as AI systems increasingly rely on professionally produced content.
Speaking at the Digiday Publishing Summit in Miami, Adam Greenberg, vice president of strategic partnerships at The New York Times, outlined three principles the publisher uses when evaluating potential AI licensing partnerships: a meaningful value exchange, a sustainable relationship and control over how its content is used.
The approach comes at a time when publishers and AI companies are trying to establish a commercial framework for using news content in AI products, search experiences and model development.
For publishers, the question is no longer simply whether AI companies will use their content. Increasingly, it is about under what conditions that use should happen.
The Three Principles Behind NYT’s AI Licensing Strategy
According to Greenberg, The New York Times evaluates AI licensing opportunities through three core principles.
1. There Has to Be a Clear Value Exchange
The Times wants AI companies and technology platforms to have an agreement or express permission before using its content.
That means licensing isn’t being treated simply as a technical arrangement for providing data. It is a commercial relationship in which the publisher expects its intellectual property to have an identifiable value.
This distinction has become increasingly important as AI companies build products that can summarize, retrieve and generate information from vast quantities of online material.
The legal debate surrounding AI training remains unresolved. The New York Times is currently involved in litigation against OpenAI and Microsoft over the use of its journalism in AI development, while the companies have disputed the Times’ allegations and argued that their use of copyrighted material is protected by fair use.
That legal dispute exists alongside a growing market for voluntary licensing agreements.
2. The Partnership Needs to Be Sustainable
The Times also does not want licensing to become a simple one-time transaction.
Greenberg told Digiday that the publisher wants partnerships to involve ongoing collaboration, rather than simply accepting a lump-sum payment and ending the relationship.
That could include developing new products, improving how journalism reaches audiences or creating mechanisms that benefit publishers beyond the individual agreement.
This is significant because AI products are changing quickly.
A licensing contract negotiated around today’s AI search or chatbot environment could look very different from what publishers need several years from now.
Greenberg described AI content marketplaces as still underdeveloped and said it could take years for systems capable of dynamically valuing content for AI training or inference to mature.
For publishers, that creates a difficult pricing problem.
How much is a piece of journalism worth when it might be used for training, retrieval, summarization, inference or a future product that does not yet exist?
There is no universally accepted answer.
3. Publishers Need Control and Guardrails
The third principle is control.
The Times wants conditions governing how and where its content can be used.
That issue goes beyond copyright ownership. Publishers also have to consider attribution, user experience, brand identity, audience relationships and whether AI-generated answers reduce visits to their own websites.
This is particularly important for a subscription publisher such as The New York Times.
The company’s business model depends heavily on building direct relationships with readers rather than simply maximizing the distribution of individual articles.
Greenberg said the Times does not want an AI agreement that interferes with its ability to build those direct audience relationships and daily reader habits.
That helps explain why referrals and links are part of the conversation.
Links and Attribution Are Becoming Part of the Deal
AI licensing agreements could give publishers an opportunity to negotiate how their journalism is presented inside AI products.
Greenberg said partnerships can provide publishers with more influence over attribution, linking and referral mechanisms that send users back to their websites.
This matters because AI interfaces can potentially answer a question without requiring users to visit the original article.
For a traditional publisher, that creates a fundamental tension:
The content can reach more people while the publisher potentially receives fewer direct visits.
The New York Times’ Amazon agreement illustrates one possible approach.
Announced in May 2025, the multiyear agreement allows Amazon to use content from The New York Times, NYT Cooking and The Athletic for AI-related experiences, including summaries and short excerpts in Amazon products and training Amazon’s proprietary foundation models. The announcement also specifically referenced direct links to Times products.
The financial terms of the publicly announced agreement were not disclosed.
That deal is important because it demonstrates that licensing can be structured around more than simply giving an AI company access to content. Distribution and referral can also form part of the relationship.
Why the Amazon Deal Fits the Times’ Framework
Greenberg specifically pointed to the Amazon relationship as an example of a partnership that satisfies the Times’ three principles.
The agreement provides a commercial value exchange, establishes an ongoing relationship and includes conditions governing how the publisher’s content is used.
The Times’ decision is notable because the publisher has taken a much harder position toward unauthorized AI use.
In September 2026, Reuters reported that the ongoing New York Times–OpenAI litigation is focused in part on whether OpenAI and Microsoft can rely on the fair-use doctrine for using copyrighted material to train AI systems.
The contrast is instructive.
The Times is not necessarily treating AI companies as a single category. Instead, it appears to be distinguishing between authorized commercial relationships and disputed or unauthorized uses.
The Bigger Problem: How Do You Price Content for AI?
The Times’ position also highlights one of the hardest unresolved issues in the AI publishing economy.
Traditional digital publishing has relatively familiar metrics.
A publisher can measure:
- Page views
- Subscribers
- Advertising impressions
- Referral traffic
- Engagement
- Conversions
AI changes the equation.
An article could contribute information to an AI-generated answer without generating a traditional page view.
A model could use content during training without sending a reader directly to the publisher.
A search assistant could retrieve information from multiple sources and synthesize it into a single response.
That makes attribution and economic value much harder to measure.
Google, for example, has recently been testing a separate AI licensing model that reportedly pays publishers based on how significantly their content contributes to AI-generated responses across products such as Gemini, AI Overviews and AI Mode.
These different approaches suggest that the industry is still experimenting with what an AI-era content marketplace should actually look like.
Why This Matters Beyond The New York Times
The Times is one of the world’s largest subscription publishers, but the underlying questions apply to news organizations of many sizes.
Publishers increasingly need to decide:
Should AI companies license our content?
If the answer is yes:
What should they pay?
And beyond payment:
How should the publisher’s content be attributed, linked, displayed and controlled?
Those questions become even more important as AI systems become a larger gateway to information.
For publishers, the goal may not simply be to extract licensing revenue. The more strategic objective could be maintaining a relationship with the audience while making sure AI platforms do not turn valuable journalism into an invisible commodity.
That is why the Times’ three-part framework—value, sustainability and control—is worth watching.
The AI licensing market is still developing. The deals being negotiated today may ultimately help establish the commercial rules that publishers and AI companies follow for years to come.
FAQ
What does The New York Times look for in AI licensing deals?
The Times has identified three core principles: an appropriate value exchange, a sustainable long-term partnership and control over how its content is used.
Does The New York Times license its content to AI companies?
Yes. The publisher has entered into at least one publicly announced AI-related licensing agreement, a multiyear deal with Amazon announced in May 2025. The agreement covers content from The New York Times, NYT Cooking and The Athletic.
Why are AI licensing deals important for publishers?
AI systems can summarize and synthesize information without necessarily sending users to the original publisher. Licensing agreements can provide publishers with compensation while allowing them to negotiate conditions around attribution, linking, content use and audience relationships.
How much did Amazon pay The New York Times?
The financial terms were not disclosed in the original announcement. Later reporting by The Wall Street Journal, as summarized by TheWrap, put the reported annual value at approximately $20 million to $25 million, but those figures were not publicly confirmed by the parties.
Is AI content licensing the same as AI copyright law?
No. Licensing is a commercial arrangement in which a rights holder permits specified uses of content. Copyright litigation concerns whether particular uses of copyrighted works are legally permissible without permission. The ongoing New York Times–OpenAI litigation illustrates that distinction.
Source Note
This QueryFeedz story is based primarily on the September 23, 2026 Digiday report on Adam Greenberg’s comments at the Digiday Publishing Summit. The key claims were cross-checked against reporting on the New York Times–Amazon licensing agreement, Reuters’ coverage of the ongoing copyright litigation, and OpenAI’s published position on the case. The article uses the source reporting as research rather than reproducing its structure or wording.
