Review AI-generated documentation PRs
Use this workflow to run the main DraftView promise: the human review layer for AI-generated docs.
Why this workflow matters
- AI text can sound correct while containing factual errors.
- Diff-only review often misses wording and procedural issues.
- Teams need proof that qualified humans reviewed output before merge.
Outcomes this workflow gives your team
- Keep up with AI output volume through a queue and lifecycle.
- Run doc-native review that non-technical experts can complete.
- Close the feedback loop by verifying what the AI changed.
- Prove oversight with sign-off and governance records.
Steps
Filter queue items by state and authorship, then open items in New, In review, Addressed, or Ready to merge.
Review the rendered docs and leave comments or edits where content is wrong, unclear, or incomplete.
Submit feedback so the author or agent can respond in GitHub.
After a new commit, verify whether it touched and fixed your commented regions.
A maintainer or admin signs off. When CI is green, move to Ready to merge and merge in GitHub.
How addressed detection works
DraftView marks feedback as Addressed when the follow-up commit changes the commented region. If a commit changes unrelated sections, DraftView keeps feedback open.
Handle stalled items
If an item stays in Waiting on AI without response, DraftView can move it to Stalled. Maintainers and admins can nudge it back to Waiting on AI.