Why AI Makes Things Up, and How to Check Its Answers
AI chatbots sometimes state wrong facts with full confidence. Why it happens, when it matters most, and a simple routine to check an answer.
AI chatbots are impressive, but they sometimes give answers that sound right and are completely wrong. This is often called a hallucination.
Why it happens
A chatbot does not look facts up the way you search a library. It predicts the words most likely to come next, based on patterns it learned. Usually that produces a good answer. Sometimes it produces something that sounds right but is not true: a made-up statistic, a wrong date or a book that does not exist.
When it matters most
Be extra careful with:
- Health, legal and money questions
- Numbers and dates
- Quotes and sources ("Who said this?" "Which study shows this?")
- Recent news, which the AI may not know about
A simple checking routine
- Ask for sources, then open them yourself. If a source does not exist, do not trust the answer.
- Check with a second source, a trusted website or a second AI tool.
- Ask it to explain its reasoning. "Show me how you got this number."
- Ask a real expert for anything important: a doctor, a lawyer, an accountant.
Treat AI like a smart friend
Think of AI as a clever friend who is helpful but sometimes wrong. You would listen to their advice and still check it before acting on anything important.
Download the free guide Research and fact-checking with AI to practise.
