AI 'hallucinations', explained in plain English

You ask an AI a simple question and it answers instantly, fluently, with total confidence. The only problem is that it’s wrong, and it invented the whole thing. That is a hallucination, and it’s one of the most important things to understand about how these tools actually work.

In 60 seconds

A hallucination is when an AI states something false as if it were true. Not a typo or a misheard word, but a confident, well-written, completely invented answer. It happens because of what language models really do under the hood: they predict plausible text, not verified facts. Most of the time, plausible and true point in the same direction. When they don’t, you get a hallucination.

The plain-English version

Picture a very well-read student who is graded only on how fluent and confident they sound, never on whether they’re actually right. Ask about something they know, and they nail it. Ask about something they don’t, and instead of admitting the gap, they produce a smooth, believable answer that happens to be fiction.

A language model works a bit like that student. It is built to produce the most plausible next words, and “I’m not sure” often sounds less plausible than a confident guess. So it guesses, in complete sentences, with no visible sign that it left the map.

Why it matters now

As people lean on AI for real work like research, legal drafting, medical questions, and code, a confident wrong answer is far more dangerous than an obvious one, because it looks right. The most common fix is to stop letting the model answer from memory alone and instead ground it in real sources it can quote, which is exactly the idea behind RAG.

What people get wrong

• “A hallucination means the AI is broken.” It isn’t a glitch you can fully switch off. It’s a side effect of how the model works, so the goal is to manage it, not assume it’s gone.
• “It only happens with obscure questions.” It can show up anywhere, including fake citations, invented quotes, and references to sources that don’t exist.
• “Newer models don’t hallucinate.” They do it less, but none are immune. Treat confidence as a writing style, not as proof.

The one-line takeaway

A hallucination is an AI confidently making something up, because it’s built to sound plausible rather than to know. The habit that protects you is simple: verify anything that matters, especially when the answer sounds perfect.