Hallucination.
When an LLM creates an inaccurate, false, or fabricated output. LLMs are prediction machines, trained on a huge body of past data (their corpus). Hallucinations can happen because of gaps or errors in that training data, but more often they're just a side effect of the prediction process itself. It does this confidently (even when wrong), and even when challenged directly it can take several turns of pushing before a model will admit something was fabricated or inaccurate. The major labs claim to have brought hallucination rates down for simple tasks, but it remains a huge challenge for generative AI, which is stochastic by nature.
In the wild
“most of the sources appear to be LLM hallucinations”
Seen here: The Guardian, 2026
For AI users
Double-check everything going somewhere important, including maths and numbers. Generative AI will present guesses as fact and may insist it's right, sometimes for several exchanges before it admits it was fabricated. Use your custom instructions to allow the model to say "I don't know" rather than fabricate as a first port of call, it's not fool proof but it does reduce hallucinations.
For business leaders
AI has been known to fabricate, and in 2026 it became an expensive lesson for PwC, KPMG, EY, and Deloitte when they were all separately caught publishing reports with fabricated content including citations. AI can be a powerful research tool but as with any source it's best to take a risk-based approach and create ways to check outputs especially if they're going somewhere mission-critical. Don't rely on vendors' standard error rates they publish in marketing materials alone as a guide for hallucination rate.