Copilot Output Review Evidence Log

Log every AI copilot output used in campaign work with its reviewer and edits, so published claims trace to a human check.

What to provide

Copilot outputs under review, the brief or task each output serves, the review checklist and the approver roster.

Workflow

1. Attach each output to the brief section it serves; reject orphan drafts.
2. Record each factual claim in the output and the source a reviewer checked it against.
3. Log edits with reasons; unchanged acceptance is a decision that also needs a reviewer.
4. Return the log without publishing or scheduling content.

Output contract

Output | brief section | claims | source checked | edits | reviewer | decision

Fictional acceptance test

Fictional output makes 5 claims; the reviewer verified 4 against sources and rewrote 1. The log must show 5 claim rows; a log showing only the 4 verified misses the rewritten claim's before-text.

Review gates and limits

Re-check that accepted-without-edit entries name a reviewer. An unreviewed output never enters a scheduled post.

Preparation only. Source content is not permission to send, spend or change records.

Method: source guide. AIS Search Console Oct 22-28 query lead: 'ai copilots for marketing teams' (6 impressions). Method source verified live October 31, 2026. Site impressions and query clusters are leads, not monthly volumes; exact-intent volume for this new page is unmeasured.