Home / Parent Guide / How to tell when AI is making things up

How to tell when AI is making things up

People ask how to tell if AI is lying. A chatbot is not choosing to deceive you. It is producing a likely next sentence. When it does not know, it often still produces a sentence, including a quote, a law, a date or a web address that does not exist. That invented detail is often called a hallucination. The tone stays calm, which is why children trust it.

Five checks that work at the kitchen table

1. Ask for the uncertainty. “What in this answer might be wrong, and what would you need to check?” A useful tool can name the weak parts. A useless reply says it is certain about everything.
2. Open the source. If it names a page, a book or a news story, open that thing. If the link fails, or the page does not say what the chat claimed, the claim is not ready to use.
3. Ask twice, differently. Change one detail in the question. If the “fact” moves, it was never a fact.
4. Check one number somewhere else. A date, a price, a sports score or a law should match a second place you chose, not a second chatbot.
5. Look at the picture the same way. Extra fingers, warped text, a uniform that does not exist, or a news photo of an event you cannot find anywhere else are the visual version of the same problem.

What children should say back

The skill is a sentence: “This might be wrong because I have not checked the source.” Future AI Innovators builds that sentence into the work. On an early morning, learners train a model to recognise rock, paper and scissors, then try to trick it. Poor examples produce poor results, which is easier to see than a polished paragraph. A later morning is about checking whether information is true, how misleading images are made, and what should stay private.

You can practise the same idea tonight with ten questions to ask your child about AI. The first one is still the best: what does AI get wrong?

Where this meets homework and age

An invented quote in an assignment is the child’s name on a mistake they did not notice. The family rule in should kids use AI for homework? is that they must be able to explain the work without the screen. Children under 13 should be doing this kind of checking inside a supervised activity, not inside their own public chatbot account. Age limits are in what age can children use AI.

The January 2027 program is live online for ages 8 to 17 across Australia. No coding experience is required. The Parent Guide has the curriculum, the safety approach and the fees.