Hands on AI education for kidsAI sounds just as sure when it is wrong as when it is right. AI literacy means knowing how it works well enough to spot the difference, and being willing to say “that is wrong”.
Spot the made-up answer
Round 1 of 4. One of these is made up. Tap the one you think is fake.
Both answers always sound the same. That is exactly what makes AI mistakes hard to spot.
You do not need to build AI to understand it. Four ideas are enough to know when to trust it and when to check.
Not magic, and not a mind. It predicts what comes next, using what people gave it to learn from.
When it does not know, it can still give an answer that sounds perfect. Nothing in its voice changes.
Exact facts, numbers, sources and recent news are where to look twice.
Disagreeing with something that sounds so sure takes a little courage. That courage is part of the skill.
1997
the year it got a name
“Misuse refers to over-reliance on automation, which can result in failures of monitoring or decision biases.”
Decades before today’s chatbots, researchers studying pilots and others working with automatic machines noticed this: when a machine is usually right, people stop checking it, and its mistakes become theirs. The fix is a habit anyone can learn.
Parasuraman & Riley, Human Factors, 1997
Scientists put AI to the test. Here is what they caught.
55%
ChatGPT-3.5
18%
ChatGPT-4
Researchers asked ChatGPT for short reports with references, then checked all 636 of them. That is how many were invented: titles that looked real but were never written. Newer versions made up far fewer, but still some.
Clue: Check that a source actually exists.
Walters & Wilder, Scientific Reports, 2023
People read AI answers to quiz questions and guessed how likely each was to be right. They thought the AI was right more often than it was, and longer explanations made them more sure, even when the answer was no better.
Clue: Long and confident is not the same as correct.
Steyvers et al., Nature Machine Intelligence, 2025
Newest doctors
Most experienced doctors
27 doctors read X-rays with help from an AI that researchers set to give some wrong answers. When it was wrong, doctors at every experience level often went along with it.
Clue: Knowing a lot is not enough. Checking is what protects you.
Dratsch et al., Radiology, 2023
A little know-how goes a long way.
+26.5%
A short set of tips on spotting false news made a nationally representative group of US adults 26.5% better at telling real headlines from fake ones, and it still showed weeks later. A highly educated online group in India improved by 17.5%, though a rural group with little social media use did not change.
Guess et al., PNAS, 2020
Honest AI helps
When AI explanations were changed to match how sure the AI really was, people got better at telling its right answers from its wrong ones.
Steyvers et al., Nature Machine Intelligence, 2025
The picture of AI gets more detailed as kids grow. The habit of checking stays the same.
Level 1
Young children can understand that AI makes guesses, and that guesses can be wrong even when they sound grown-up.
Try asking“Is that a fact, or is it guessing? How could we check?”
Level 2
Kids love testing AI with questions they already know the answer to. Finding a mistake themselves is the lesson that sticks.
Try asking“Ask it something you know really well. Did it get it right?”
Level 3
Older kids can learn where AI slips most — exact numbers, quotes, sources and recent news — and how to check them somewhere else.
Try asking“Where would you look to check that is actually true?”
Five quick questions for any AI answer. Tick them off as you go.
Trust everything and you pick up its mistakes. Trust nothing and you miss a useful tool. AI literacy is the middle: knowing how it works, where it slips, and feeling able to say so.