Hallucination
When AI states something false with total confidence. Its mistakes sound exactly as sure as its correct answers.
In practice
Inventing a report from an institute that does not exist.
What an AI hallucination is
An AI hallucination is an answer that is false but delivered as if it were true. The AI might invent a statistic, a quote, a book title, a research study or a website, and present it in the same calm, detailed tone it uses for facts it gets right.
The word is a little misleading. The AI is not seeing things. Chatbots work by predicting the most likely next words based on patterns in what they learned from. Most of the time the likely answer is also the correct one. When it is not, the AI has no built-in alarm that goes off, so it carries on writing something that sounds right.
What hallucinations look like
Common examples: a citation to a paper that does not exist, a quote attributed to a real person who never said it, a confident date or number that is slightly off, a summary of a book that mixes in events from a different book, or a link to a page that was never there.
The details are what make them convincing. A made-up report usually comes with a plausible title, an institute name and a year. That level of detail is exactly why a student who skims the answer will believe it.
How to catch one
Reading the answer more carefully does not help, because a hallucination reads exactly like a correct answer. What works is checking the claims that matter against a source you can actually open: a textbook, an official website, the original article. If the AI names a source, look that source up yourself. If you cannot find it, treat the claim as unconfirmed.
Asking the same AI whether it was right is not a check. It can apologize and then invent a second wrong answer just as confidently. Names, numbers, dates, quotes and anything medical, legal or about a real person are the claims to check first.
Why it matters for kids
Children often treat a fluent answer as an authoritative one, and AI is very fluent. Learning that confidence is not evidence is one of the most protective habits a young AI user can build. It also changes how they use AI for homework: as a starting point to verify, not a finished answer to copy.
Questions parents ask about this
Where students meet it
Hallucination is taught in Lesson 13, Trust But Verify: AI Hallucinations & Fact-Checking, part of the Ethics & Trust unit of the AI Explorers course for kids. Soleil is the guide for that lesson, and it is one of 8 terms in the unit.
More from Ethics & Trust
Back to all 54 AI terms for students.
Knowing the word is the start
Hallucination is one of 54 terms students meet across 16 video lessons that teach them to direct AI rather than just use it.
