In plain English
A hallucination is when the assistant states something false with total confidence. An invented quote, a made-up study, a feature a product doesn’t actually have, a date that’s just wrong. It doesn’t hedge, it doesn’t flag it, it just says it like fact.
It happens because the model is predicting plausible-sounding text, not looking anything up. Unless it’s got a search tool or a file switched on, it’s working from patterns, not a database it’s checking against. So it can produce something that reads perfectly confident and is simply untrue.
A worked example
Ask an assistant for “a quote from a famous economist about inflation” with no source given, and there’s a real chance you get a quote that sounds right but was never actually said. It’s not lying exactly, it’s just filling the shape of “a quote an economist might say”.
Now give it the actual article or transcript and ask it to quote only from that, or ask it to say when it’s unsure rather than guess. Both cut the risk right down. Anything that actually matters, though, still needs checking by you.
The usual mix-up
People assume a hallucination means the assistant is broken, or worse, that it’s lying to them on purpose. It’s neither. It’s doing exactly what it always does, predicting the next likely bit of text, and it has no idea it’s wrong. There’s no dishonesty in it because there’s no awareness in it. That’s kind of the whole point.
Where you’ll use it
- How to get AI tostop making things upResearch & discover · free tools
- How to get AI tofact-check somethingResearch & discover · free tools
- How to get AI tobe honest with meResearch & discover · free tools
Also comes up in: automate repetitive tasks, help me research a topic, update itself.