AI Hallucination

An AI hallucination is a response that is plausible and well-formed but not grounded in fact or in the provided source - an invented citation, a wrong figure, a made-up API. It is a direct consequence of how large language models work: they predict likely text, not verified truth, so fluency is no guarantee of accuracy.

Hallucination is usually mitigated with retrieval-augmented generation to ground answers in real sources, with human review of high-stakes output, and with narrower, verified generation. In a governed system it is also a reason to constrain what a model is allowed to assert or act on, and to keep a record of what it produced - so a wrong output can be traced rather than merely regretted.