This issue was meant to be an update on bilateral licensing deals. We expected plenty of new additions to the list from April 2026 (75 deals) completed between platforms and publishers over the last six months. Well, this is a bit awkward, but the big licensing deals have slowed significantly. At an event in London we heard from one publishing executive that talks with AI companies have “all but stopped”. The UK collective licence (PLS) has 250 publishers signed up, but as far as the conversations with the platforms go, they are still ongoing. 

Meanwhile the crawlers are busier than ever. An FT executive described one trial account trying to pull a million articles in eight hours. So is the licensing market dead? No. But as it shifts, it is sending useful signals about what the platforms actually want today. And it is certainly not what they wanted two years ago. 

Meta: a land grab at face value, but a deeper shift behind it 

In October Meta rewrote the terms that every business account on Facebook and Instagram agrees to. Buried inside: permission for Meta’s “scrapers, bots, crawlers and user agents” to roam any website “associated with” a business, create test credentials, and use what they find. No negotiation, no price, and no opt-out short of deleting the account. The first draft even said “train”. 

Meta’s explanation is that the changes consolidate existing ad terms and were never aimed at publishers. The timing offers a clue to what Meta was aiming at. It launched Muse, its shopping agent, on 8 September. Amazon blocked it on 20 September for turning up without permission. Three weeks later, terms appeared that would hand Meta that same permission from every business on its platform, and a lot of them are merchants. The clause reads like it was drafted for its next-generation agentic commerce product.  

But intent is not the point. A licence obtained through the small print is still a licence, and it sets the market price for that publisher’s content at zero. That is a threat to every licensing opportunity a publisher has, whether or not Meta ever builds a news product with it. 

And this time, the industry reflex worked. Publisher bodies on both sides of the Atlantic pushed back within days. Meta delayed, then paused, then promised consultation. On 30 October the CMA opened a formal investigation, and the PPA is feeding it member evidence. The ask is simple: separate access to the platform from consent to use copyright content. That is a systematic solution, and one that applies across platforms. 

Google: paying for supply, keeping the customer 

Google’s News AI pilot and its “400 publications signed up” headline sound like a licensing boom. They aren’t. This is a boom in publishers looking for new ways to monetise their content, and Google obliging with new formats to distribute it to audiences. The pilot is a flat fee for supplying Google News with content for AI previews and audio briefings: a distribution deal with an AI badge on it. It could be a useful revenue stream for a publisher, but it is not a licensing deal. 

The actual licensing product is the other one. The contribution pilot, the scheme that touches search answers, pays around 100 publishers by a formula nobody outside Google can see. Google is buying supply cheaply and keeping the audience relationship for itself. For Google, that is a better deal than negotiating bilateral, and possibly expensive, agreements. 

The AI labs: shopping somewhere else 

OpenAI and Anthropic haven’t lost interest in data, but they have moved their budgets. Each now spends around a billion dollars a year on expert human data, paying doctors, lawyers and engineers $100-plus an hour to grade and build training tasks. Anthropic is the biggest buyer of reinforcement learning environments and has no publisher deals at all. OpenAI buys whole websites as training gyms for its agents. They are not queuing up for another news archive. Most archives have already been read. 

Why this is good news 

Because the fog is lifting. Two years ago the market was one product (a training licence) sold to one kind of buyer at a price negotiated behind closed doors. Now there are at least five ways for a publisher to get paid: flat supply deals, pay-per-crawl, pay-per-use meters, collective licences, and whatever Google’s contribution pilot grows into. We know which platform wants what. And we know where regulation will and won’t step in: the CMA will police coercive terms and enforce separation, protecting publishers from the unfair use of platform power over copyrighted material. 

It is also clearer where the leverage sits for publishers: 

• Retrieval (consumer and lifestyle titles with fast-moving content), not archives: fresh, structured, metered content is what is still being bought. 

• Expertise (B2B, professional and specialist publishers), not just text: a verified specialist archive and a contributor base of practitioners is exactly what the labs pay most for. 

• The reader and the transaction (any title where affiliate or commerce is a meaningful revenue line, which for many lifestyle brands is a fifth or more of digital): an agent will happily read a publisher’s review and buy elsewhere, unless the publisher is the one manning the checkout.