AI Hallucination
When an AI model generates information that sounds confident and plausible but is factually incorrect, fabricated, or not supported by its sources.
What Is AI Hallucination?
An AI hallucination occurs when a language model produces output that is fluent and authoritative-sounding but factually wrong, invented, or unsupported by any real source. The model is not lying in a human sense — it has no concept of truth. It predicts the most statistically likely next token given the context, and sometimes the most likely-sounding continuation is simply false.
Hallucinations take several forms: factual hallucinations (inventing a statistic, date, or quote), citation hallucinations (fabricating a paper, URL, or legal case that does not exist), contextual hallucinations (contradicting information provided in the prompt), and reasoning hallucinations (producing a confident but logically flawed chain of steps). They arise because LLMs compress vast training data into weights and fill gaps with plausible patterns rather than retrieving verified facts.
Why It Matters
Hallucination is the single biggest barrier to deploying AI in high-stakes domains such as medicine, law, and finance. A model that is right 95% of the time but confidently wrong 5% of the time — with no signal indicating which is which — is dangerous in any context where errors carry real consequences. Understanding why hallucinations happen, and how to mitigate them with grounding techniques like RAG, citations, and human review, is essential for anyone building trustworthy AI products.