An AI hallucination is when AI makes something up and presents it as fact. It isn't lying on purpose. It is guessing, and the guess comes out sounding real.
Hallucinations can look like:
- A quote no one ever said
- A book, study, or news article that doesn't exist
- A website link that leads nowhere
- A phone number that belongs to someone else, or to no one
- A law, rule, or medicine dose that sounds official but is wrong
This happens less than it used to, especially with chatbots that search the web, but it still happens. And because the made-up part is mixed in with true information, it can be hard to spot.
How to spot one
- Check that the source exists. Search for the book, study, or article by its exact title. If you can only find it in the chatbot's answer, be suspicious.
- Search quotes in quotation marks. If the exact words don't show up anywhere else, the quote may be invented.
- Never use a phone number or link from AI for your bank, Medicare, or any company. Get it from your card, a bill, or the official website you typed in yourself.
- Ask the chatbot directly: "Are you sure? Please show me where this comes from." Sometimes it will correct itself.
Try this
- Ask a chatbot for "three books about the history of my town." Then look each one up at your library's website. Are they all real?
- Ask a librarian. Helping people find real sources is what librarians do best. In Oregon, Answerland (answerland.org) offers free help by chat, in English and Spanish.