What is a content licensing deal?
A content licensing deal is a contract that lets an AI company use content it does not own, on agreed terms. The content owner might be a news publisher, a forum, a stock image library or a book publisher. The AI company might want the content to train its models, to quote and summarize it inside AI answers, or to get faster, structured access to new material.
A typical example: on May 22, 2024, OpenAI agreed with News Corp to integrate news content from The Wall Street Journal, the New York Post, The Times and The Sunday Times into its AI platform [2]. Another: on February 22, 2024, Google announced an expanded partnership with Reddit that gave Google access to Reddit's Data API, described as "real-time, structured, unique content" [1].
The term is used broadly. Some "deals" are pure training licenses, some are mainly about display and attribution, and some are revenue-sharing programs that publishers join rather than negotiate.
What do AI content licensing deals usually include?
AI content licensing deals usually include one or more of four things: permission to train on the content, permission to show it in answers, structured or real-time access to it, and some form of compensation. The mix varies by deal, and most deals do not publish their full terms.
| Deal element | What the AI company gets | Example |
|---|---|---|
| Training rights | Permission to use archives to train models | Often part of deals, terms rarely disclosed |
| Display rights | Permission to quote, summarize and link to content in answers | OpenAI and News Corp [2] |
| Data access | A structured feed or API rather than crawling | Google and Reddit Data API [1] |
| Revenue share | Publishers share in ad revenue earned next to their content | Perplexity Publishers' Program [3] |
The Google and Reddit announcement also included a cloud element: Reddit agreed to use Google's Vertex AI to improve search and other features on Reddit [1].
When did AI content licensing deals start?
AI content licensing deals took off in 2023 and 2024, as AI companies faced copyright lawsuits and wanted reliable, fresh content for chat products. The wave included platforms, news groups and, by mid-2024, revenue-share programs open to many publishers.
2023: lawsuits set the stage
Authors filed copyright suits against OpenAI from July 2023, and The New York Times sued OpenAI in late December 2023 [2]. Lawsuits and licensing deals have moved in parallel ever since: some publishers sign, others sue.
2024: the first wave of deals
Google announced its expanded Reddit partnership on February 22, 2024 [1]. In May 2024, OpenAI signed deals with News Corp, Axios and Vox Media [2]. In July 2024, Perplexity launched its Publishers' Program to share advertising revenue with partner publishers, after accusations that it had reproduced publishers' work without permission [3].
2025: courts and settlements
In June 2025, a US federal court ruled in Bartz v. Anthropic that training its models on the authors' books was fair use, but that its use of more than seven million pirated library copies was not, and Anthropic agreed a settlement with the authors in September 2025 [4]. Publishers continued to sue as well: Ziff Davis sued OpenAI on April 24, 2025 [2], and Perplexity faced suits from publishers including Dow Jones [3].
What is the difference between a content licensing deal and opting out of AI training?
A content licensing deal grants an AI company permission on negotiated terms, while opting out withholds permission without any agreement. Opting out is one-sided and usually free; licensing needs both parties to agree.
Google's main opt-out is Google-Extended, a robots.txt product token. It controls whether content Google crawls may be used to train future Gemini models and for grounding in Gemini Apps and Vertex AI [5]. Google states that Google-Extended "does not impact a site's inclusion in Google Search nor is it used as a ranking signal" [5]. Blocking other AI crawlers, such as OpenAI's GPTBot, works the same way for those companies.
What is the difference between a licensing deal and a copyright settlement?
A licensing deal sets terms for future use by agreement, while a copyright settlement resolves claims about past use. The Bartz v. Anthropic settlement in September 2025 was paid to authors to end a lawsuit over books already used, not to buy rights to new content [4]. The two are often reported together, but a settlement does not create an ongoing relationship the way a license does.
Does a content licensing deal help a site appear in AI answers?
There is no published evidence that a content licensing deal raises how often a publisher is cited in AI answers, and the AI companies have not said that licensed sources rank higher. Deals that include display rights give an AI company permission to show the content, but none has published how that affects which sources get cited for a given question.
Google's position is notable. Announcing the Reddit partnership, Google said it "does not change Google's use of publicly available, crawlable content for indexing, training, or display in Google products" [1]. And Google says its AI features rely on its core Search ranking systems to retrieve pages [6]. In other words, Google's AI answers draw from the open web through normal ranking; a license is not described as a shortcut.
Two things are worth separating:
- Access. A deal can give an AI company content it could not otherwise reach, such as paywalled archives or a real-time feed. That content may then appear more often simply because it is available.
- Preference. Whether licensed content is deliberately favored over equally relevant unlicensed content is not publicly documented for any major AI system.
What should site owners do about content licensing deals?
Most site owners will never be offered a content licensing deal, so the practical question is how to stay visible in AI answers without one. Licensing deals have so far been struck with large publishers and platforms with big archives or unique, fresh content.
For everyone else:
- Decide your crawler policy deliberately. Blocking AI training crawlers is a business choice. Know which tokens control training and which affect search, since Google-Extended does not change Google Search inclusion [5].
- Earn visibility on the sites that are licensed. If AI assistants lean on a forum, news group or review site, being discussed and cited there carries your brand into those answers.
- Keep your own pages strong in normal search. Google's AI features retrieve through core ranking systems [6], so ordinary SEO remains the route into Google's AI answers.
Treat claims that a deal guarantees AI visibility with caution: no AI company has published data to support it.