What is an AI hallucination and why does it happen?

An AI hallucination is when a language model confidently generates information that sounds plausible but is completely made up or factually wrong. It's not a glitch in the traditional sense — it's a byproduct of how these models work. They're prediction engines, not databases. When you ask a question, the AI isn't retrieving a stored fact. It's calculating the most probable next word in a sequence, one token at a time, based on patterns it learned during training. Sometimes the most statistically likely word string creates a convincing fiction. I've seen this happen with biographical details especially. Ask for a CEO's background and you might get a flawless paragraph about their Harvard MBA and early career at McKinsey — except the person never attended either. The model isn't lying. It's just pattern-matching the shape of a typical CEO bio. This matters because hallucinations can be dangerously subtle. A medical query might return a drug dosage that looks correct but is off by a decimal point. A legal question might cite a court case that never existed. Google's 2024 AI Principles documentation acknowledges this as an unresolved challenge across all large language models. The practical takeaway: treat AI outputs as first drafts or suggestions, never as verified facts. For anything where accuracy matters — health, finance, legal — always cross-check with a primary source. For a deeper dive into getting reliable outputs, see our guide on how to write AI prompts that reduce these errors.