Fact-Checking
1 min read · Reviewed 17 Oct 2026
In short
Verifying specific claims outside the tool that produced them.
In plain English
Checking the things a model states, using something other than the model. Names, numbers, dates and citations are where errors concentrate; vague claims rarely cause trouble.
A real example
A model gives you three sources. You search for each one. Two exist and say roughly what was claimed; one does not exist at all. That third one would have gone out in a client report.
Why it matters
The tone of an AI answer carries no information about its accuracy. Verification is the only signal you get, and it takes minutes.
