AI 101

Hallucination

Also known as: hallucinate · confabulation

Basics

A hallucination is a confident, fluent statement that is false or invented. The model isn’t “lying” — it’s a next-token-predictor, and sometimes the most likely-sounding continuation is just not true.

Common shapes:

  • Inventing facts, dates, or citations.
  • Claiming a capability it doesn’t have.
  • Confidently misquoting a source.

RAG and citations exist largely to reduce them by grounding answers in real documents.

Working with it: treat model output as *a draft to verify*, especially for facts, numbers, and quotes. Never paste unverified model output into something that ships.