Hallucination
When an AI states something false or invented as if it were true, with examples and practical ways to reduce the risk.
Hallucination: when an AI model produces information that sounds convincing but is false or invented, such as a made-up statistic, quote, source or link.
In plain English
A language model writes what is likely to come next, not what has been checked. When it lacks the right information, it can still produce a fluent answer that is simply wrong.
Common examples
- A citation to a study or article that does not exist.
- A quote attributed to a person who never said it.
- A wrong date, number or name stated with confidence.
- A link or product feature that does not exist.
How to reduce the risk
- Give the AI source text and tell it to use only that text.
- Ask it to say “I don’t know” or “not in the source” instead of guessing.
- Ask for sources, then open them yourself.
- Re-ask in a fresh chat and compare answers.
- Use tools that search the web, but still check the results, because they can misread sources too.
- Have a qualified person check legal, medical, tax and financial answers.
Common confusion
- Hallucinations are not lies. The model has no intent, it is predicting text.
- A confident tone says nothing about accuracy.
- Newer models hallucinate less in many cases, but not never.
Related terms
LLM, RAG, temperature.