Home / Guides / How AI works, in plain words

What Is an AI "Hallucination"?

A hallucination is an answer that sounds right but is false. It is one of the most important things to understand about AI.

Step by step

  1. Know it happens.
  2. Ask for sources.
  3. Open the sources.
  4. Check key facts elsewhere.
  5. Be extra careful with names, dates and quotes.

What to expect

Expect fake book titles, wrong dates, invented quotes or non-existent web links. They often sound polished.

Mistakes to avoid

Do not use an AI quote or citation without finding the original. Avoid asking for obscure facts, which are most likely to be invented.

A good next step

Ask for sources, open them, and ask AI to say when it is unsure.

Why it happens

  • AI predicts likely words. It does not look facts up the way a person checks a book.
  • When it does not know, it may still produce a confident answer.
  • Made-up sources, quotes and numbers are common examples.

Protect yourself, step by step

  1. Ask for sources. Then open them and check that they exist and say what is claimed.
  2. Ask it to say when it is unsure. "If you do not know, say so."
  3. Double check anything important. Health, legal, money, dates and names.
  4. Ask the same thing a different way. Different answers are a warning sign.
  5. Use a search tool for current facts.

Example

  1. What happened. Someone asked for a book on a topic and got a title and author that did not exist.
  2. The lesson. Search the title before you trust it.

A prompt you can copy

Type this into ChatGPT, Claude or Gemini

Explain what an AI hallucination is with an example, and give me five habits for catching them.

Good to know. If the answer matters, verify it.

Want to try this with a friendly teacher?

Join a free live class on Zoom. Small group, no experience needed, no sign-up fee.

Reserve a free seat

Common questions

Why does it do this?

It predicts likely words, not truth.

Is it getting better?

Yes, but it still happens.

Can it be fixed?

Models are improving, but errors still happen.

Is it lying?

No. It does not know it is wrong.

Should I stop using AI?

No. Use it for drafts and ideas, and verify facts.

More guides