The publisher AI licensing market has settled into two tracks: big individual deals and consortium models. Neither currently guarantees fair value for most publishers.
Publishers are being asked to license the content that makes AI useful. The question is whether the terms reflect what that content is actually worth.
The publisher AI licensing market is settling into shape. It doesn't look particularly good for most publishers.
Two structural models have emerged. The first is large individual deals: News Corp signed a multi-year agreement with Meta in March for up to $50 million a year, giving Meta access to News Corp archive content across its US and UK titles. These deals are significant but exclusive to publishers large enough to negotiate individually.
The second model is consortium-based: the News/Media Alliance signed a licensing deal with AI startup Bria that lets its 2,200 member publishers opt in to monetising retrieval augmented generation (RAG) demand. Publishers get paid based on how often their content gets used by Bria's enterprise clients.
Publisher members of the News/Media Alliance now able to monetise AI licensing through the Bria consortium deal
A Digiday publisher scorecard on Big Tech's AI licensing terms reflects the frustration. Most publishers rate the terms as below fair value. The complaint is structural: AI companies need publisher content to train and operate models, but the pricing reflects a power asymmetry, with AI companies setting terms and publishers choosing to accept or walk away.
The statutory licensing argument is gaining momentum globally. Policymakers in Europe, Indonesia, Latin America and at the World Intellectual Property Organization are exploring laws that would require AI companies to pay publishers as a class, not just those large enough to negotiate individually.
RAG-based compensation models like the Bria deal are more structurally sound than one-time training data deals. They create recurring revenue tied to ongoing usage. But the per-query rates need to be validated at scale before publishers can assess whether the economics actually protect their business.
The window to negotiate is narrowing. AI models are already trained on historical content. The ongoing leverage is in live, regularly updated content that makes RAG-based responses accurate and trustworthy. Publishers who wait may find their negotiating position has weakened as AI companies build proprietary content pipelines.