Open Editor
Privacy
8 min read

AI Content Licensing Deals in 2026: Who's Actually Getting Paid, and How Much

A handful of publishers are collecting nine-figure checks from OpenAI, Google, and Amazon. No publicly disclosed deal has ever paid less than eight figures — which is really a story about who the deals leave out.

DealReported Value
News Corp → OpenAI$250M+ over 5 years (~$50M/yr)
Reddit → Google~$60M/yr (part of $203M total disclosed in IPO filing)
News Corp → Meta~$50M/yr
NYT → Amazon$20–25M/yr
Financial Times → OpenAI$5–10M/yr

The dollar figures behind AI content licensing deals are, for the most part, public knowledge — reported by the outlets involved, disclosed in IPO filings, or leaked to trade press. What's less often stated plainly is what those numbers actually reveal: this is a market with a floor, and the floor sits somewhere around eight figures. Below that line, a publisher hasn't been priced out of a bad deal — they typically haven't been offered a deal at all.

Tracking how a licensing announcement's language shifts between drafts? ClearDiff makes side-by-side comparison instant.

The Headline Deals: Real Money, Concentrated Among a Few Names

OpenAI has signed roughly two dozen publisher and data-licensing agreements since mid-2023, with partners including the Associated Press, Axel Springer, the Financial Times, News Corp, Vox Media, Condé Nast, Reddit, and Stack Overflow. The largest reported figure is News Corp's, at more than $250 million spread across five years — described by one industry analysis as roughly 2.5 times News Corp's five-year net income, which gives some sense of how material these payments are even to a company that size. Google's biggest disclosed deal runs through Reddit, reportedly worth around $60 million annually, part of the $203 million in aggregate data-licensing contract value Reddit disclosed in its own IPO filing. Amazon's deal with the New York Times, covering Alexa and Alexa-based shopping features, is reported at $20–25 million a year.

The Real Ceiling: No Disclosed Deal Under $10 Million

This is the detail that reframes the whole market: as of 2026, no publicly disclosed AI content-licensing deal has ever been reported below roughly $10 million. That's not evidence such deals don't exist — private terms could fall below that line without being reported — but it's a reasonable signal about where the negotiating threshold sits. Deal value tracks negotiating leverage far more than content quality: leverage that comes from owning a large, well-known archive and being able to credibly threaten litigation or public criticism if a company trains on that archive without paying. A mid-sized regional publisher, a niche trade outlet, or an independent blog — regardless of how good the writing is — generally doesn't have either of those levers to pull.

The headline number in most coverage of AI licensing is how much the biggest publishers are getting paid. The more consequential number is the one that doesn't appear in any of these reports at all: what everyone below the eight-figure threshold gets, which historically has been nothing.

Anthropic Took a Completely Different Path

Anthropic is a useful contrast precisely because it hasn't signed a single publisher licensing deal of this kind. Instead, facing a class-action copyright suit from authors, it agreed to a $1.5 billion settlement in September 2025 — the largest publicly reported figure in this entire space, but a legal settlement rather than a forward-looking content license. It compensates for past use rather than establishing an ongoing commercial relationship, which is a meaningfully different thing than what News Corp or Reddit signed. The distinction matters for anyone trying to compare "how much AI companies are paying for content" across companies — Anthropic's number and OpenAI's numbers aren't actually measuring the same kind of transaction, a nuance covered in more detail in the earlier post on the NYT v. OpenAI litigation, where the choice between licensing and litigating is playing out in real time.

Two New 2026 Programs Aimed at Everyone Else

Two developments this year specifically target the publishers locked out of the eight-figure club, using a fundamentally different payment structure: revenue share instead of a flat fee.

Perplexity formalized its Comet Plus program in January 2026, allocating a $42.5 million pool funded by a $5/month subscription tier, with participating publishers keeping 80% of the revenue their citations generate. More than 2,400 publishers have enrolled — an entry threshold and structure that looks nothing like a nine-figure lump-sum negotiation, open to outlets with no litigation leverage at all.

Google's move is newer still: a "AI contribution pilot," rolling out via a Search Console dashboard widget, that pays publishers when their content is judged to have meaningfully contributed to an AI-generated response across Gemini, AI Overviews, and AI Mode. There's no upfront fee, payment is usage-based, and publishers can opt out at any time. According to reporting on the rollout, the program has specifically been offered to dozens of small and mid-sized publishers rather than the large names that already hold direct licensing deals — the opposite end of the market from where the headline dollar figures live.

ModelWho QualifiesPayment Structure
Flat-fee licensing (OpenAI, Google, Amazon direct deals)Large publishers with negotiating leverageLump sum or annual fee, $10M+ reported floor
Perplexity Comet PlusAny enrolled publisher (2,400+)80/20 revenue share from a $5/mo subscription pool
Google AI contribution pilotSmall and mid-sized publishers (invite-based)Usage-based, no upfront fee, opt-out anytime

Why the Same Deal Gets Reported With Different Numbers

Anyone comparing multiple sources on these deals will notice the figures don't always match exactly — Reddit's Google deal shows up as "$60 million a year" in some reporting and "$65–70 million a year" in others. This isn't necessarily one outlet being wrong; multi-year contracts with performance-based components can reasonably be annualized or totaled in different ways depending on what a given report is measuring, and companies rarely confirm exact figures on the record. Treating any single reported number as precise to the dollar is less reliable than treating it as a well-sourced estimate within a range.

What This Means If a Publisher Isn't News Corp

For the overwhelming majority of sites publishing text online, the realistic 2026 options aren't a nine-figure OpenAI deal — they're robots.txt-based blocking of AI crawlers, enrollment in a revenue-share program like Comet Plus where eligible, or acceptance into a usage-based pilot like Google's, none of which come close to the disclosed value of the handful of marquee deals. That's a real shift from where things stood even a year earlier, when the only options were a direct deal (available to almost no one) or no compensation at all. It's not parity with what News Corp negotiated, but it's the first meaningful entry point smaller publishers have had into this specific revenue category.


The honest summary: a small number of large publishers are collecting real, disclosed, eight-and-nine-figure sums from OpenAI, Google, Amazon, and Meta — and no smaller deal has ever been publicly reported, which strongly suggests the market simply hasn't extended that far down. Anthropic sidestepped the licensing question entirely with a legal settlement instead. The two genuinely new developments in 2026 — Perplexity's revenue-share model and Google's usage-based pilot — are the first real attempt to build a payment structure for the publishers who were never going to get a phone call from a licensing negotiator in the first place.

For questions or inquiries contact us at info@cleartexteditor.com