A confident answer is not a source.
Why fluent AI output needs an evidence check.
Documented failure mode · Updated 2026-09-27
The distinction
NIST describes confabulation as confidently presented false or erroneous output. This can include fabricated references. Readability does not establish reliability.
A habit worth building
Before relying on an important answer, open its original sources and check whether they support the actual claim. A real publication title alone is not enough. If verification fails, leave the question unresolved rather than filling the gap with confidence.
Keep the context in view
Our practical suggestion: make the depth of verification proportional to the consequence of being wrong. Use AI to help investigate a question; keep the responsibility for accepting an answer explicit.
SOURCE & FURTHER READINGNIST · Generative Artificial Intelligence Profile, 2024