What is AI Hallucination?

An AI hallucination is an output from an AI model that sounds confident but is factually wrong or entirely made up. A language model may cite a paper that does not exist, invent a legal precedent, or state a wrong number in a fluent, plausible sentence.

Hallucinations happen because generative models predict likely text rather than retrieve verified facts. The risk grows when the model has no access to reliable source data, when its training data is outdated or inconsistent, or when a question falls outside what its data covered.

Common mitigations include retrieval augmented generation, grounding answers in curated sources, and human review for high-stakes output. In enterprise settings, the data side matters as much as the model: when teams can trace which data state an answer was based on, wrong outputs become easier to catch, compare, and reproduce.

Related terms: Ground Truth · Retrieval-Augmented Generation (RAG) · AI Reliability Gap · Verifiable Data State

Frequently asked questions

Why do AI models hallucinate?

Generative models predict plausible text rather than look up verified facts, so gaps, conflicts, or outdated records in their data can produce confident but wrong answers.

Can AI hallucinations be prevented?

They can be reduced rather than fully eliminated. Grounding models in curated data, retrieval augmented generation, and human review of high-stakes outputs all lower the rate.

Are hallucinations a model problem or a data problem?

Both. The model generates the wrong text, but missing, outdated, or untraceable data is often what makes the wrong answer possible and hard to catch.