Staleness
Also called Staleness
When the information a system serves has quietly fallen out of date.
Think of it like
A wall calendar still showing last month — right where you left it, wrong about today.
Example
A support bot keeps telling users to click a button that was renamed two releases ago, because its docs index was never rebuilt.
How it actually works
Staleness is the failure that freshness measures. It’s dangerous precisely because stale answers look confident and correct — nothing flags them as old. It creeps in when source data changes but the index, cache, or embeddings don’t. The fix is disciplined refresh and invalidation; the trap is caching aggressively for speed and forgetting to expire it.
For product teams
The silent source of "the AI told me something that used to be true" complaints.
For engineers
Occurs when source mutations aren’t propagated to derived indexes/caches; mitigate with TTLs and change-driven invalidation.
Related
- Freshness — The metric that catches it.
- Index Refresh — The routine that clears it.
- Semantic Cache — Caches are a common culprit.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome