An AI “hallucination” is a confident answer that is false or unsupported — an invented statistic, a book that doesn’t exist, or a legal case that was never decided. Knowing why it happens makes it easier to catch.
Why it happens
Language models generate text that is statistically likely to follow your request. They are optimised to sound fluent and helpful, not to guarantee truth. When information is missing, ambiguous or outside their training data, they may fill the gap with something plausible.
Warning signs
- Very specific numbers or dates with no source.
- Citations, links or quotes you cannot find elsewhere.
- Answers about very recent events.
- Different answers when you ask the same question again.
A quick verification routine
- Ask the assistant for its sources, then open and read them yourself.
- Cross-check key facts with an official or reputable source.
- Search for exact titles or quotes to confirm they exist.
- For health, legal or financial matters, consult a qualified professional.
Rule of thumb
The more an answer matters, the more you should verify it. Use AI to speed up research — not to replace it.
