Proportionate optimisation, control 7 of 8
O-07 Require independent contextual pre-publication review
P0Hard gateNeeds your evidence
An optimiser cannot be the sole source of its release verdict. A trusted reviewer must inspect the candidate, rationale, context and evidence outside the authoring agent's editable approval boundary.
- Priority
- Gate. A FAIL stops publish or launch.
- Hard gate
- Yes. A FAIL stops publication, whatever the score.
- Change class
- READ_ONLY: Inspect authorised evidence.
- Quality bucket
- Not scored: hard gate only
Evidence to review
- Trusted reviewer identity, dated evidence/artifact references, candidate/action binding and six completed contextual assessments.
How to implement
- Use six review areas: purpose, naturalness, site intent, winner preservation, business truth and policy integrity.
- Bind the verdict to site, environment, action payload, baseline, candidate and affected cohort.
- Separate semantic review from deterministic tests; each must identify what it did and did not evaluate.
- Accept a separate model-assisted review only within trusted identity and enforcement boundaries; a second chat or display name alone is not independence.
- No available independent reviewer means HOLD the consequential action, not silently self-approve.
- Preserve disagreements and resolved false positives; an owner cannot override a confirmed prohibited practice through an ordinary approval.
How UltimateSEO checks this
Needs your evidence. No automated check covers this control, so it stays UNKNOWN until evidence is supplied.
States used: PASS, FAIL, UNKNOWN, NOT_APPLICABLE or BLOCKED. UNKNOWN is never shown as a pass. How states work.
Part of The UltimateSEO Framework v4.3, catalogue dated 2026-10-14. Evidence states, control states and the publish gate are explained on the hub.