Start here
Configure AI attribution
Choose attribution signals and understand known, estimated, mixed, and unknown work.
- Audience
- owners, administrators, engineering leaders
- Required role
- Owner or administrator.
- Product access
- Measure
- Navigation path
- Measure > Settings > AI attribution
Outcome
Configure attribution inputs and interpret AI Share without overstating uncertain authorship.
Before you begin
- An active repository
- A clear internal policy for any self-reported AI signals
Choose attribution settings
Open AI attribution settings and enable only signals your team actually uses. Keep unknown work visible rather than forcing every line into AI or human categories.
- 1
Review available signals
Confirm which provider metadata, commit markers, and repository evidence are available.
- 2
Apply organization policy
Enable supported inputs and document any developer behavior required to produce them.
- 3
Save and reprocess if needed
Apply changes to future analysis, and use the offered reprocessing action when historical results should be updated.
Review confidence
Known means direct evidence exists. Estimated means FeatureFactory inferred attribution from supported signals. Mixed means the change contains more than one attribution state. Unknown means evidence is insufficient.
Validate the result
Open AI Share, then inspect several pull requests from teams with known tool behavior. Compare the summary with the evidence on each pull request before using the trend in a review.
Protect honest reporting
Do not interpret unknown as human-authored or estimated as certain. Report confidence beside adoption and compare AI Share with quality and delivery outcomes.