Since late 2023, AI companies and publishers have moved from arguing about scraping to signing contracts. OpenAI has announced partnerships with news groups, Google has a data deal with Reddit, Perplexity has launched a revenue-sharing subscription, and Microsoft has opened a marketplace for licensing content to AI products. For publishers, the question is no longer whether these deals exist but whether a particular one is worth signing.
This guide summarises publicly announced deals and the main licensing models, using the companies' own announcements and dated reporting. It then sets out what publishers give up and gain, and a list of questions to ask before you sign anything.
One caution up front: most financial terms are confidential, and reported numbers are often estimates. We only describe what was announced or reported by named outlets, and this is not legal advice.
Key Takeaways
- Licensing deals usually bundle several rights: access to current and archived content, display of summaries in AI answers, and sometimes use for model training.
- Models now range from flat-fee contracts and revenue shares to usage-based marketplaces, pay per crawl and machine-readable licensing standards.
- Training rights and display rights are different things; negotiate them separately where you can.
- Crawler controls (GPTBot, OAI-SearchBot, Google-Extended and others) are your baseline leverage, whether or not you sign.
- Ask about attribution, links, usage reporting, exclusivity and exit terms before price.
Why AI companies license content
AI assistants need two kinds of content: large volumes of text to train models, and fresh, reliable information to ground answers about current events. Publishers hold both, including archives and paywalled journalism that crawlers cannot freely reach. Licensing gives AI firms legal clarity and access; it gives publishers payment, attribution and some say over how their work is used. It also shapes how often AI tools cite and link to a publisher, which matters as zero-click searches grow.
Publicly announced deals and programmes
| Deal or programme | Announced | What was made public |
|---|---|---|
| OpenAI and Axel Springer | 13 Dec 2023 | ChatGPT summaries of selected stories from brands such as POLITICO, Business Insider, BILD and WELT, including paywalled content, with attribution and links; content also used to help train OpenAI's models |
| Google and Reddit | 22 Feb 2024 | Google gets access to Reddit's Data API for real-time structured content; Reddit uses Vertex AI. No amount in Google's announcement |
| OpenAI and News Corp | 22 May 2024 | Access to current and archived content from mastheads including The Wall Street Journal, The Times and The Australian, shown in response to user questions; no amount in OpenAI's announcement |
| Cloudflare pay per crawl | 1 Jul 2025 | Publishers set a per-request price; crawlers pay, get an HTTP 402 price response, or are blocked. Launched in private beta |
| Perplexity Comet Plus | Aug 2025 | Reported by Axios: $5 per month subscription, 80% of revenue to publishers, $42.5 million set aside for early partners |
| RSL 1.0 (Really Simple Licensing) | 10 Dec 2025 | Open standard for machine-readable AI usage and compensation terms; backers include Cloudflare, Akamai, Creative Commons and the IAB Tech Lab |
| Microsoft Publisher Content Marketplace | 3 Feb 2026 | Reported by Search Engine Land: publishers set licensing terms, AI builders license content for grounding, with usage-based reporting; co-designed with publishers including the AP, Condé Nast, Hearst and Vox Media |
OpenAI's News Corp announcement also shows how narrowly rights can be drawn:
“The partnership does not include access to content from any of News Corp’s other businesses.”
— OpenAI, OpenAI: News Corp and OpenAI sign landmark multi-year global partnership (22 May 2024)
For a closer look at one programme aimed at a wider range of publishers, see our guide to the Perplexity publisher programme.
The main licensing models
| Model | How it pays | Best for | Watch out for |
|---|---|---|---|
| Flat-fee contract | Fixed annual or multi-year fee | Large publishers with archives and strong brands | Fee may not track how much your content is actually used |
| Revenue share | Share of a product's subscription or ad revenue | Publishers willing to bet on a product's growth | Payouts depend on adoption you cannot control |
| Usage-based marketplace | Payment linked to how content is used for grounding | Publishers wanting transparency and reporting | Early-stage; rates and demand still forming |
| Pay per crawl | Price per request from AI crawlers | Sites with infrastructure partners that support it | Crawlers may choose not to pay and simply not access you |
| Standard licence terms (RSL) | Publishes machine-readable terms crawlers can honour | Smaller publishers without negotiating power | Relies on AI companies choosing to respect the terms |
| No deal | None; you choose which crawlers to allow | Publishers prioritising control or awaiting better terms | Possible loss of visibility and referrals in some AI products |
What publishers trade off
Training rights versus display rights
Allowing a model to learn from your archive is a one-way door: once trained, the knowledge cannot easily be removed. Allowing an assistant to display summaries with links is ongoing and can be renegotiated. The Axel Springer deal covered both. Where possible, price and time-limit them separately.
Paywall value
Summaries of paywalled stories can expose readers to your brand, or can satisfy them without a subscription. Ask how much text is shown, whether links point to the paywalled page, and whether you receive data on referrals. Our guide to paywalled content in AI search covers the technical side.
Traffic and attribution
A deal that brings payment but few clicks may still be worthwhile, but you should know which it is. Independent research points the same way: Pew Research Center found in 2025 that Google users clicked traditional results less often when an AI summary appeared, so set up AI traffic tracking in GA4 before and after any agreement.
Leverage and control
Your crawler settings remain your baseline. OpenAI documents GPTBot (training) separately from OAI-SearchBot (ChatGPT search). Google documents Google-Extended as a control over use in Gemini model training and grounding, which Google says does not affect inclusion or ranking in Google Search. See our guides to AI crawlers, Google-Extended and Cloudflare AI Crawl Control.
Questions to ask before signing
- What exactly is licensed? Current content, archive, paywalled content, images, video? Which brands or sections are excluded?
- What uses are allowed? Training, grounding answers, displaying summaries, building new products? Can each be switched off?
- How is attribution handled? Will answers name your publication and link to the original page?
- What reporting will you receive? Usage counts, referrals, top content, regions?
- How is payment calculated and audited? Fixed, revenue share or usage-based, and how can you verify it?
- Is there exclusivity? Can you sign with other AI companies too?
- How long does it last, and what happens at the end? Must trained models be retrained or can they keep what they learned?
- How are errors handled? Is there a route to correct misattributed or inaccurate summaries of your journalism?
- What happens to crawling? Does the deal require you to allow specific crawlers, and on what terms?
Options for smaller publishers
Headline deals have mostly gone to large groups. Smaller and niche publishers still have choices:
- Join open programmes or marketplaces where eligible, and compare their reporting.
- Publish machine-readable licensing terms using a standard such as RSL.
- Use infrastructure-level controls to allow, charge or block crawlers.
- Focus on being cited: original reporting, data and expertise remain what AI systems cannot easily get elsewhere. Our guide for publishers in AI search covers this.
Common mistakes
- Signing a broad training licence without separately valuing it.
- Accepting a deal without usage or referral reporting, leaving you unable to judge its value.
- Blocking every AI crawler in protest, then losing visibility in search-style AI products you depend on.
- Relying on reported deal figures from other publishers as if they were benchmarks; most terms are confidential.
- Agreeing to exclusivity without understanding what competitors are being offered.
- Forgetting that contributors, photographers and syndication partners may hold rights you cannot license.
Related Guides
- Making the Business Case for AI Search to Leadership
- How to Choose an AI Search (GEO) Agency: Questions and Red Flags
- Pricing AI Search Services: Cost Drivers and Engagement Models
Frequently Asked Questions
Do AI licensing deals include model training?
Some do. OpenAI's Axel Springer announcement said content would help train its models, while other deals focus on displaying content in answers. Check each contract for separate training and display rights.
How much do AI companies pay publishers?
Most deals do not disclose financial terms. Some figures have been reported by media outlets, but they are not reliable benchmarks for other publishers.
Can I license content without a direct contract?
Options include revenue-share programmes, usage-based marketplaces, pay per crawl through infrastructure providers, and publishing machine-readable licence terms with standards such as RSL.
Will blocking AI crawlers improve my negotiating position?
It can signal that access has value, but it may also remove you from AI search products. Decide crawler by crawler, since training and search crawlers are often separate.
Does Google-Extended stop my content appearing in AI Overviews?
Google says Google-Extended controls use for Gemini model training and grounding in some other Google systems and does not affect Google Search inclusion. For Search features, Google points to controls such as nosnippet and max-snippet.
Conclusion
AI licensing is becoming a real revenue line for some publishers, but the deals differ widely in what they buy and what they give back. Decide what you are willing to license, measure your current AI traffic, and use crawler controls deliberately while you negotiate. If you want help measuring how AI products use and cite your content, or protecting traffic as AI answers grow, see how our SEO services can help with AI traffic loss, or get a quote.
References
- OpenAI: Partnership with Axel Springer (13 December 2023)
- OpenAI: News Corp and OpenAI sign landmark multi-year global partnership (22 May 2024)
- Google, The Keyword: An expanded partnership with Reddit (22 February 2024)
- Axios: Perplexity says it will give publishers 80% of Comet Plus revenue (26 August 2025)
- Cloudflare blog: Introducing pay per crawl (1 July 2025)
- RSL Collective: RSL 1.0 specification released (10 December 2025)
- Search Engine Land: Microsoft launches Publisher Content Marketplace for AI licensing (3 February 2026)
- Google Search Central: Google's common crawlers (Google-Extended)
