Why does AI “hallucinate”?
“Hallucination” is the common name for an answer that sounds convincing but contains made-up or false claims.
What is it?
A “hallucination” is an answer that sounds convincing but contains made-up or false claims. It happens because a language model assembles likely text and has no built-in fact check. Errors are more common with rare information, numbers, quotes and sources.
An everyday example
A student asks a chatbot for three scientific papers on a topic and gets neatly formatted citations, one of which does not exist.
Why does it matter?
Invented sources and figures can end up in essays, news articles and even in court. That is why checking is essential.
A common misconception
That an answer with a cited source has been checked. You have to open the source and check that it exists and says what the model claims.
In more detail
Is the model lying on purpose?
We cannot infer intent from a wrong answer. Technically, the important point is that the system does not always check whether its output is true.
What can it make up?
Book titles, court rulings, quotes, calculations, events and even web links that look completely genuine.
What helps most?
Primary sources, using the right tools, a clearly limited context and human review, especially for important decisions.
Will the problem disappear completely?
Nobody can guarantee that today. The error rate depends on the model, the task, the data and how answers are checked.
Remember: If AI cites a source, open it. A cited source is not yet a checked source.
Reliable sources
Last reviewed 1 November 2026. An explainer, not professional, legal or medical advice. For rules and figures that change, check the official version.