Skip to content
Bot Tamers

How to teach kids to fact-check AI

By the Bot Tamers team. Last updated .

The most useful AI skill a child can learn is also the simplest: do not take the answer on trust. Here is why AI gets things wrong, a short routine kids can repeat, and an activity you can do together in about ten minutes.

Why AI can sound sure and still be wrong

Common Sense Media's guide to using AI for school explains it plainly: AI tools "can sound confident and still be wrong," and "AI predicts what sounds like a good answer rather than truly understanding the question." It also warns that "AI can 'fill in the gaps' with people, events, or sources that look real but aren't."

That last point surprises many parents. A chatbot can name a study, an expert or a book that does not exist, in the same calm tone it uses for true facts. The tone tells you nothing about whether it is right.

For a child, one sentence is enough: AI picks words that sound right, so it can make things up, and it does not warn you when it does.

A check routine kids can repeat

Common Sense advises parents to "encourage your child to double-check AI's answers," and for schoolwork to "always verify with class notes, textbooks, or teacher-approved sites." HealthyChildren.org, from the American Academy of Pediatrics, suggests asking kids where information comes from, whether the source has an agenda, and why they should trust it or not.

Put together, that makes three questions a child of 8 to 12 can remember:

  1. Who says so? Is there a real source, and can I find it?
  2. Can I find it somewhere I trust? A textbook, a teacher-approved site, or a grown-up who knows.
  3. Does every part check out? One true fact does not make the whole answer true.

A 10-minute activity: spot the mistake

This builds on an idea from Ying Xu, who studies AI in learning at Harvard, in an interview with UNICEF: type a question into a chatbot with your child, look at the answer together, and ask "Which parts seem confusing or not quite right?"

  1. Pick a topic your child knows well. A favorite animal, a sport, a place you have visited. You do the typing.
  2. Ask an AI tool for five facts about it. Read them aloud together.
  3. Guess first. For each fact, your child says "sure", "not sure" or "I think that's wrong," before checking anything.
  4. Check each one. Use a book, an encyclopedia, or a site you trust. Mark each fact true, false, or could not confirm.
  5. Talk about it. Did any wrong fact sound just as sure as the right ones? Did the AI name a source? Could you find it? Xu suggests ending with: "How might the tool have created that response?"

If every fact checks out, that is a good result too. The habit is to check, not to expect a mistake every time.

Pictures and videos too

The same habit applies to images, video and voices, which AI can also make. The American Academy of Child and Adolescent Psychiatry (AACAP) notes that deepfake technology "can be used to create new forms of cyberbullying." HealthyChildren.org suggests making a game of telling real from fake. The question to teach is the same: who posted this, and can I find it from a source I trust?

How Bot Tamers practices this

Checking AI is one of the four skill areas in the Bot Tamers curriculum. In the ages 8 to 10 lesson "Checking what AI tells you," Bit writes a short report about the Moon that credits a made-up claim to a scientist who does not exist. Your child looks up each claim in a pretend trusted source and finds the one that does not hold up. In the ages 11 to 12 unit "AI gets it wrong," kids fact-check an AI-written report, learn the warning signs of made-up answers, and see that some mistakes have no warning signs at all.

There is no live chatbot: Bit is scripted, so every mistake is placed on purpose for your child to catch. Our safety page explains what else that means for your child's privacy. For the bigger picture, read what AI literacy means for kids 8 to 12 or how to talk to your child about AI.

Bot Tamers runs as an iOS app for families and is not on the App Store yet. Pricing will be: 7 days free, then $9.99/month or $59.99/year. Renews automatically until canceled. Cancel anytime in iPhone Settings at least 24 hours before the trial ends. To hear when a spot opens, join the beta waitlist.

Sources