RAG & Search
Retrieval quality
How well a retrieval system returns the information actually needed to answer a question.
Example
A test set checks whether search consistently retrieves the passage needed for a correct answer.
Why it matters
It helps teams identify why a source-backed assistant misses information people need.
You'll hear it
“The writing is fine; search just isn't bringing back the right evidence.”
Can you control it?
Set by the team
The people building or running the AI choose this setup. An everyday user generally needs their help to change how this part works.
Common questions
- Can a good-looking final answer hide poor retrieval?
- Yes. The AI may produce convincing wording even when the useful source was missed.
- Can empty results be the correct outcome?
- Yes. If no relevant material exists, returning nothing is better than treating unrelated text as evidence.
- Does quality stay fixed after documents change?
- No. Additions, removals and outdated records can change what searches return.