AI Explorers · lesson 17 of 24 · 25 min
A student asked a chatbot for three books about a local history topic. It gave three titles with authors and summaries. None of the books existed. Here is why that happens and how to protect yourself.
Hallucinations
When an AI system states something false as if it were true, people call it a hallucination. It might invent a fact, a quote, a source, a date, or a math answer, and say it with total confidence.
Why? Remember how LLMs work: they predict likely-sounding text. A book title, an author’s name and a summary have a familiar pattern. The model can produce text that fits that pattern perfectly without checking whether the book is real. It is not lying on purpose. It has no intentions at all. It is pattern-matching, and the pattern does not include “check the facts.”
When hallucinations are more likely
Be extra careful when you ask about:
- Specific facts: exact numbers, dates, names, quotes and statistics.
- Sources: book titles, article links and references.
- Local or rare topics: your neighborhood’s history, a small business, a person who is not famous. There was little text about them to learn from.
- Recent events: a model learned from text up to a certain date, and may not know what happened after, unless the tool can search the web.
- Math with many steps: some chatbots slip on long calculations.
How to fact-check AI
Use these four habits that good researchers use:
- Stop: notice when an answer contains facts you will rely on.
- Find the source: look for the information in a trustworthy place, like a library database, an official website, or a textbook.
- Cross-check: see if two or more reliable sources agree.
- Ask a person: a teacher, librarian or parent can help you judge a source.
A good rule: use AI to help you brainstorm, explain and practice. Use real sources to get facts. Never copy an AI’s “facts” into homework without checking them.
Not just facts
AI can also “make things up” in other ways. It may describe a picture with details that aren’t there, or confidently give advice that is wrong for your situation. For anything about health, safety, money or legal issues, always talk to a trusted adult or a real expert.
Some AI tools can search the web and show links to their sources. That helps, but it does not solve the problem. The AI can still misread a source or mix up two of them. Click the link and read the original yourself. If there is no source at all, treat the answer as a starting point for your own research, not as the final word.
Try it
This activity works with no AI at all.
- Write a short “fact sheet” of 5 sentences about a topic you know well, like your favorite team, a local landmark or your school.
- Secretly change 2 of the sentences so they sound believable but are false. Make the fake facts confident and specific, like an AI hallucination.
- Give the sheet to a family member or friend. Their job: find the 2 fake facts.
- Ask them how they decided which ones were fake. Did they check a source?
- Write down: which fake fact was hardest to spot, and what made it convincing?
Check yourself
What is an AI hallucination?
When an AI states something false as if it were true, like a fake fact or a made-up source.
Why do language models hallucinate?
They predict likely-sounding text and do not automatically check whether it is true.
Name a type of question where you should be extra careful with AI answers.
Exact facts, statistics, sources, recent events, local topics, or health and safety advice.
Key words
- Hallucination: a false statement an AI presents as true.
- Fact-check: confirm information using reliable sources.
- Source: where information comes from.
- Reliable: trustworthy and accurate.
Previous lesson: How chatbots predict the next word
Back to AI Explorers
Next lesson: Prompting basics and staying safe