The 7 Deal Points of AI Content Licensing Agreements
I’ve tracked 44 AI/Publisher deals —these 7 terms show up again and again, with real $$$, evolving tactics, and surprising clauses
The Big AI companies—OpenAI, Google, Meta —are cutting licensing deals fast.
I’ve tracked 44. Seven key deal points keep surfacing.
For each deal point, I share:
Real examples (with $$$ where available)
Notable quotes
Shifts in how these deals are evolving
Here’s a sampling of some of the deals I track in chronological order:
…on to the 7 deal points:
1. 💰 Payment Structures
Most AI content licensing deals include two parts: a fixed upfront payment and a variable payout based on usage.
Fixed Upfront / Minimum Payment
Publishers want guaranteed cash—they’re licensing crown jewels.
Reuters: Estimated $25M one-time fee (plus another $40M spread over 3 quarters), possibly from Meta:
“…$25,000,000 of generative AI related transactional content licensing revenue.”
Source: Mike Eastwood, CFO via Transcript via Investing.com
My estimate of the other $40M is from 3 previous Reuters earnings calls in which they report lumpy bumps of $18M to $36M (most of which I think is AI).
Dotdash Meredith: $16M/year minimum from OpenAI:
“If you look at Q3 of 2024, licensing revenue was up about $4.1 million year over year... the lion’s share... driven by the OpenAI license.” —Chris Halpin, IAC COO (on earnings call)
Informa’s Taylor & Francis: $10M upfront and recurring payments through 2027 (from Microsoft):
“…both an initial data access fee ($10m+) and a recurring payment…” -Informa
Variable Payment
These fluctuate based on usage levels or performance metrics. They reward publishers whose content proves especially valuable.
Axel Springer’s $25M deal with OpenAI has a one-off payment just for training and variable back-end fees.
But the variable sounds bigger:
“…the larger fee will be paid under an annual licence agreement”
Source: Financial Times).
TIME’s variable revenue also sounds bigger:
"Some of it (revenue) is fixed, a lot of it is variable…"
– Mark Howard, COO of TIME (Source: Digital Context Next)
Credits
Some deals include more than just cash.
“The deal could be worth more than $250 million over five years, including compensation in the form of cash and credits for use of OpenAI technology…”
Source: Wall Street Journal
These credits are presumably for buying extra licenses of ChatGPT or its APIs (see Section 4 (“Publisher Access to AI Tools”) below).
2. Content Scope: What AI Companies Want
Training Rights
Training Was Central in Early Deals
Shutterstock, Stack Overflow, Reddit, Wiley, Axel Springer, Associated Press, News Corp, TIME, and others confirmed that their content is being used to train AI models.
These deals often include archives or bulk content, perfect for foundational model training.
Example: Shutterstock reportedly earned over $100M from training deals with multiple LLMs (OpenAI, Meta, Amazon, Google, Apple).
Training Value on the Decline?
When it comes to training rights, two of the most recent high-profile deals don’t follow the old playbook:
The Washington Post
Deal emphasizes summaries, quotes, and links shown inside ChatGPT.
No mention of training rights in the announcement.
🛡️ The Guardian
Deal centers on real-time content access and news visibility in ChatGPT.
Training rights are not referenced.
Display Rights
Display rights allow AI platforms to show article summaries, quotes, logos, and links within tools like ChatGPT.
Atlantic and AP content now appears directly in ChatGPT responses.
News Corp reportedly requested a delay before its content went live, but that doesn’t appear to be the case based on my tests:
Media Types & Carve-Outs
AI companies are expanding beyond text. Many deals now include rights to:
Images
Video (especially long-form libraries)
Audio and music
UGC (user-generated content like Reddit threads)
Example:
Curiosity Stream projects $19.6M in AI licensing revenue for 2025, likely tied to its 210,000-hour factual video library. That contributed to its first profitable quarter. Source: The Top 10 Media Companies Ranked by AI Revenue
Some publishers exclude specific content assets from deals.
News Corp’s $250M OpenAI deal excluded Factiva (itself an aggregator of news sources and using now using AI) and HarperCollins.
Brand Carve-Outs
Some publishers exclude brands as part of their deal.
For example, News Corp.’s $250M deal excluded Factiva (which itself is an aggregator of 30,000 content sources) and HarperCollins.
3. 🔗 “Display” Rights
Display deals give AI platforms the right to show article summaries, quotes, and logos, with links back to the publisher’s site.
These deals are designed to drive traffic and ensure attribution.
Examples of Display Language
Washington Post (the most recent big deal)
“ChatGPT will display summaries, quotes, and links to original reporting from The Post in response to relevant questions.”
—Washington Post
Attribution and Branding
Attribution terms vary:
It's easier when a single publisher's content answers a prompt directly.
More complex when results pull from blended sources.
Recent deals emphasize:
Logo placement within AI products
Direct links to original content
Joint PR and co-marketing opportunities.
4. 🛠️ Publisher Access to AI Tools & Tech
Some publishers now negotiate access to AI tools and infrastructure as part of their licensing deals.
This turns licensing into a product-development partnership—not just a content transaction.
These deals often include:
Early access to proprietary models or APIs
Joint experimentation and feedback loops
Custom tooling for newsrooms and editorial teams
Here are some examples:
5. ⚙️ Technical & Operational Requirements
Beyond payments, the tech terms matter. AI companies don’t just want content—they want easy access to it, in real time.
Integration Access: The On-Ramp for LLMs
Most content deals come with a set of keys:
APIs or Bulk Dumps — LLMs need content to flow in fast. Publishers often give API access (like a pipeline) or big file dumps (like handing over a hard drive).
Live Feeds — Real-time news feeds are 🔥 right now. If your content helps LLMs stay current, that’s a big win.
Prompt Data Is the New Gold
Prompt data is hot—but two different groups want it for different reasons.
LLMs — On one side, LLMs want to license prompts from platforms like Stack Overflow.
Why? Those user questions are gold for training.
They show what real people ask, how they think, and where the model can improve.
Stack Overflow now charges LLMs for access to these prompts—and they want attribution too.
Publishers — On the flip side, publishers want prompt logs from the LLMs.
If a user asks “What’s the best diet for heart health?” and the answer pulls from Health.com, the publisher wants to know.
Some are asking for logs that show when their content is triggered—so they can monitor usage, enforce attribution, or negotiate backend payments.
So yes—everyone wants the prompts.
Because whoever understands the questions gets closer to owning the answers.
Reporting & Support: A Sneaky Revenue Lever
Some publishers are smart about the extras:
Usage Reports — to track how much content is used (and maybe trigger bonus payments).
Tech Support — API goes down? They want someone on call.
Dashboards — In big-money deals, some publishers require real-time dashboards showing usage.
These aren’t just technical terms—they’re leverage points. Content owners, don’t give them away for free.
6. 📄 Contract Terms & Clauses
AI content deals are quietly converging on a few standard terms.
Two of the most common: non-exclusivity and clauses that protect early movers.
Non-Exclusivity Is the Norm
Most deals let publishers keep their options open. They don’t lock into just one AI partner.
I see non-exclusive terms in deals from:
Shutterstock
Reddit
Axel Springer
Associated Press
Informa (Taylor & Francis gave Microsoft non-exclusive access)
The Atlantic, which can sublicense its content to other LLMs via ProRata
Bottom line: publishers are playing the field.
First-Mover Clauses = Hidden Gold
Some early deals also protect the pioneers.
Example: The Associated Press reportedly has a clause with OpenAI that lets them “reset” the deal if someone else gets better terms.
That’s smart.
These quiet clauses might be the most valuable part of early deals—but they don’t show up in the press release.
7. Term Length
AI content licensing deals range from one-time transactions to multi-year agreements. While not always disclosed, a pattern is emerging.
Examples:
2 years: Associated Press/OpenAI
2.5 years: Reddit/Google
3 years: Axios/OpenAI, Axel Springer / OpenAI
5 years: News Corp/OpenAI
6 years: Shutterstock/OpenAI
Two additional deals are publicly described only as “multi-year”:
Reuters/Meta
Shutterstock/Reka
One-Time Licensing
Some AI content deals are simple: one check, one data dump, no strings.
These lump-sum licensing deals give LLMs access to archival content—usually just for training, not real-time use.
Take Wiley, for example on their recent earnings call
“The onetime [$23 million] transaction… includes access to previously published academic and professional book content for specific use in training LLM models.” — Matthew S. Kissner; Interim CEO, John Wiley & Sons, Inc. Wiley
That’s $23M for one shot at Wiley’s archives.
These one-off deals are usually shorter-term and tightly scoped: no updates, no fresh content—just historical data for model training.
It’s a clean way for publishers to monetize their back catalog without giving up the future.
Final Takeaways for Execs on Both Sides of the Table
For Content & Media Execs:
You don’t need to say yes to training — More publishers are holding back training rights. “Display-only” is becoming a smart default.
Your archive has value—even in a one-time deal — Wiley turned old books into $23M. If you’ve got evergreen content, there’s a check waiting.
Ask for usage reports and dashboards — They’re not just operational—they’re negotiation ammo for bonuses or renewals.
Push for product access — TIME, Axios, and The Atlantic are shaping AI tools from the inside. If you're licensing content, you should shape the roadmap too.
For AI Execs:
Attribution is no longer optional — Expect to link back, show logos, and credit the source. Publishers want visibility and traffic.
One-size licensing is fading — From academic books to breaking news to user comments, content types and rights packages are diversifying fast.
Prompt logs may be part of your future — Publishers want to see when and how their content is triggered. Be ready to share prompt-level data.
Variable payments can unlock better content — If you want the good stuff, especially niche or time-sensitive content, tie payouts to performance.
The game is changing fast. Whether you're licensing content or buying it, these next 12–18 months will define who wins—and who gets left out.
Thanks for reading!
Rob Kelly
Creator & Host of Media & the Machine
p.s.: How to reach me:
If you want to reach me, the best way is to subscribe below and reply to any of my emails.
Stellar reporting