Why AI Makes Mistakes and How to Check Its Answers
AI sometimes states something false in a very confident voice. This is often called a hallucination. It is not lying. It is guessing what sounds right. A few quick checks will protect you.
Step by step
- Watch for exact details. Names, dates, quotes, phone numbers, web addresses and statistics are where mistakes hide.
- Ask for its sources, then verify them. Ask where the information comes from. If it names a source, look it up yourself. Some invented sources look real.
- Compare with a second source. Search the key fact on a trusted site such as a government, hospital or major news website.
- Ask it to double-check. Try are you sure? and please re-read your answer and list anything uncertain.
- Raise your guard for big decisions. For health, taxes, legal and large purchases, use AI to prepare questions, then ask a professional.
A prompt you can copy
Type this into ChatGPT, Claude or Gemini
Give me your answer again, and this time mark which parts you are sure about and which parts I should verify.
Good to know. Think of AI as a very quick, very well-read assistant who sometimes gets things mixed up.
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 seatCommon questions
Can I trust anything it says?
Treat it as a helpful draft. It is right a lot of the time, but verify what matters.
Do newer versions make fewer mistakes?
Generally yes, but none are perfect.