Source Change Detection

You cannot govern what you do not notice. Source change detection is the earliest signal that AI knowledge may be wrong. Detection must catch edits, moves, deletions, and metadata shifts that change eligibility without visible copy drama.

Whole-site hashing misses surgical policy edits. Surgical edits are how legal teams work. Detection must be document-granular and severity-aware.

Deleted pages and redirects are changes too. Bots citing 404s destroy trust faster than stale text.

Metadata-only changes alter eligibility without visible copy edits. Detection must watch audience tags and lifecycle fields, not only body hash.

404 and redirect events are changes. Bots citing dead links destroy trust faster than stale wording.

Hash bodies and track metadata fields per document: audience, status, effective dates. Compare CMS version IDs where available.

Classify diffs: cosmetic, structural, factual, policy-critical. Critical diffs page owners and block releases pending review.

Capture before and after snippets for high-risk templates to accelerate reviewer work.

Validate detection after CMS upgrades and template migrations. Platforms change HTML wrappers that fool naive hashers.

Replay historical exports after incidents to find blind spots and tune sensitivity.

Severity-classify diffs so cosmetic blog tweaks do not page the same way as refund eligibility edits.

Replay detection against historical CMS exports after incidents to find blind spots and tune rules.

Nintendo change detection catches Help Center URL moves after IA redesign. Broken links would have left retrieval pointing at 404 mirrors. Detection rebinds chunks to new URLs before the chatbot cites dead links.