Scholarly publishers’ fiscal 2026 results have made the AI content-licensing trend concrete in a way trade-press speculation could not: real, disclosed dollar figures, reported to investors, not projections. Wiley’s fiscal 2026 results (year ended April 30, 2026) show $49 million in AI licensing revenue and lifetime AI revenue surpassing $110 million. The pattern underneath that headline number — large publishers with big, well-cleared back catalogues capturing meaningful recurring revenue while the long tail of small and society publishers reports little or nothing comparable — is the more consequential story for the research-administration community.
Wiley: $49 million in FY2026, $110 million lifetime
In its fourth-quarter and full fiscal-year 2026 results, released via its investor newsroom, Wiley reported $49 million in AI licensing revenue for the fiscal year, up 23% on the prior year (implying roughly $40 million in fiscal 2025), and said lifetime AI revenue had surpassed $110 million since it began signing these deals. The release names two “landmark” healthcare-AI partnerships, with IQVIA and OpenEvidence, as key drivers, alongside a broader roster of corporate AI customers. Wiley’s own framing links this revenue directly to its FY2026 net-income growth, crediting AI licensing and its research segment as the two main contributors in a year where overall revenue was roughly flat.
Earlier trade coverage (Publishers Weekly, Investing.com) of the same results corroborates the $49 million and $110 million figures and the IQVIA/OpenEvidence partnerships; some secondary write-ups have circulated a lower “$44 million” figure for Wiley’s disclosed AI revenue, which does not match the number in Wiley’s own investor release and should be treated as outdated or imprecise rather than authoritative.
Taylor & Francis, Springer Nature: the deals that set the pattern
Wiley’s disclosure sits inside a wider pattern that took shape from 2024 onward and continues to generate revenue reported in 2026 results. Taylor & Francis (part of Informa) struck a non-exclusive content-and-data licensing agreement with Microsoft reported at roughly $10 million in its first year with recurring payments structured through 2027; Informa has indicated it expects total AI-related revenue across its portfolio to exceed $75 million for the year. Springer Nature’s 2024 agreement with Google, a one-time payment reported at $23 million for a defined corpus of previously published papers, functioned as an early benchmark valuation that subsequent negotiations across the industry referenced.
These figures share a structural feature: they accrue to publishers with large, well-cleared, long-running backlists and the in-house rights and legal capacity to negotiate bespoke terms with AI developers. That is a small set of firms.
Why scale and negotiating leverage decide who gets paid
Industry analysis through 2026 (Scholarly Kitchen commentary, trade and industry-newsletter coverage) converges on the same read: this market pays for scarcity and negotiating leverage, not merely for having content that could plausibly train a model. A large commercial publisher with millions of cleared, well-metadata’d full-text records and a dedicated licensing team can negotiate; a learned society running one or a handful of journals through a hosting partner typically cannot extract comparable terms on its own, even though its content is being crawled, scraped, and in some cases licensed as part of a larger aggregator deal it had limited say in. This is the “little to no benefit for mid-size and society publishers” dynamic behind the headline numbers, and it is a genuine, actively discussed structural feature of the market as of 2026, not a settled one-off complaint.
Collective licensing models — society publishers pooling content through an aggregator or collecting society to negotiate as a bloc — are the most frequently proposed remedy in industry discussion, but no comparably-sized collective deal had closed as of these disclosures. Separately, some large publishers have pursued litigation rather than licensing where they allege unauthorized use: Elsevier’s 2026 suit against Meta over AI training use has been reported by trade press (American Chemical Society’s C&EN) as being watched closely by other scholarly publishers weighing licensing versus enforcement as their primary lever.
What this means for research administrators and authors
For research-administration offices and authors, the practical questions these deals raise are less about the headline dollar figures and more about contract terms already in force. Whether a publisher can license a given article’s text for AI training turns on the copyright transfer agreement or exclusive licence the author signed at acceptance — most standard commercial-publisher agreements assign or exclusively licence rights broadly enough to cover this use without requiring a separate author opt-in, which is precisely what has drawn author-advocacy criticism of these deals. Institutions negotiating publishing agreements or reviewing author rights policies, and authors publishing under green or gold open-access routes, should treat AI-training rights as a live term to check rather than an implicit given. CASRAI’s guides on AI training data provenance, copyright, and TDM exceptions for research and on the policy landscape for generative AI in manuscripts cover the underlying rights and disclosure mechanics in more depth.
What to watch next
Three things worth tracking through the rest of 2026: whether any collective-licensing vehicle for society and small publishers actually closes a deal at meaningful scale; whether Wiley’s growth rate (23% year-on-year) holds as the initial round of large-publisher deals matures and the pool of un-licensed major backlists shrinks; and how the Elsevier v. Meta litigation, and any similar suits, resolve relative to the licensing path most large publishers have chosen instead.
