Why it matters
Verification confirms a source is accurate, authorized, and appropriate for an AI audience before indexing. Skipping verification pushes garbage into vectors at scale.
How it works
Checklist reviews: authority, audience, accuracy spot checks, conflict scan against policy, metadata completeness. Failures return to authors.
Example
Nintendo verifies new Help Center billing articles against Refund Policy before public-safe approval. Articles with conflicting refund windows fail verification automatically via contract tests.
Common mistakes
- 1Verification by non-subject-matter reviewers ticking boxes
- 2One-time verification with no re-check on source change
- 3Verifying web HTML but not linked PDF attachments
Your source changed. Your AI should know.
Track source freshness, detect changes, and review impact before answers drift.
Explore Source Maintenance