Hallucination
In AI, a hallucination is when a model states something false as though it were plainly true. It might invent a quotation, cite a study that does not exist, or confidently give the wrong date. This happens because language models are built to produce plausible sounding text, not to check facts, and they have no built-in sense of what is true. Hallucination is one of the main reasons AI answers should be verified before they are trusted.
To understand why hallucinations happen, it helps to know what a language model is actually doing. At its core, it predicts the next word over and over, choosing whatever continuation seems most likely given everything so far. It is astonishingly good at producing text that reads well and usually holds together. What it is not doing is consulting a database of established facts or asking itself “is this actually correct?” There is no little truth detector inside. If a confident, well formed falsehood fits the pattern of the conversation, the model will happily produce it.
A useful analogy is a fluent student sitting an exam who has not revised a particular topic. Rather than admit they do not know, they write a smooth, confident paragraph that sounds right and hope for the best. The prose is impeccable; the content may be nonsense. AI models do this without any intent to deceive, because generating convincing language is precisely what they were built to do.
This is why hallucination is such an important idea for anyone relying on AI. It is not a rare glitch that will soon vanish; it is a side effect of how these systems work. Techniques such as retrieval-augmented generation, where the model looks up real sources before answering, can reduce it substantially, and models are improving. Even so, the sensible rule for now is to treat AI output as a knowledgeable draft to be checked, not as a final authority, especially for anything that really matters.