The technical decision
Use a field-level freshness policy. Product coverage may need frequent review; a definition or formula may remain valid longer; an original study should keep its data window even after commentary is updated.
Implementation sequence
- Tag each claim as stable, periodically changing, or event-driven
- Store a verified date and source for every unstable product or platform fact
- Update the page only when the evidence, recommendation, or reader outcome changes
- Set lastmod to the real publication or material-update date rather than the build date
- Retest internal links, structured data, and citations after a substantive revision
A page-level example
A comparison table reviewed in August should not silently keep a January price. Either verify the row, label it unknown, or remove the field. The page date should reflect the material edit, not an automated site rebuild.
Evidence, limits, and UnderAI's role in content freshness for AI search
Keep these fields with the decision:
- URL
- template
- rendered field
- expected value
- observed value
- deployment date
- test result
- rollback owner
Frequent publishing and frequent date changes are not substitutes for accuracy. Stale facts can damage trust even when the page remains indexed.
UnderAI tracking observations are dated by platform, market, Prompt Group, and answer snapshot, which lets editorial updates distinguish a fresh measurement from a timeless claim.