Your 7-Year-Old Just Asked the Robot a Question. Here’s What to Say Next.
I asked a seven-year-old what he knew about AI.
“Something that tells you about different stuff. Like how to save somebody, and how to stop something.”
Then I asked him whether AI is real.
“The answer is true, but whoever is coming up with the answer is fake.”
I have been turning that sentence over ever since. He is seven. Nobody sat him down and explained any of this to him. He worked it out on his own, in the way children work things out, and he got half of it more right than most adults I know.
The other half is the reason I am writing this.
The half he got right
He knows there is no person in there.
This is the thing most parents brace for. That their child will start treating the assistant like a friend. That they will say goodnight to it. That somewhere underneath, a small person is forming an attachment to something that cannot form one back.
He is not confused about that at all. Whoever is coming up with the answer is fake. He has already sorted the voice from a person, without help, and filed it correctly.
If you have been worrying about that particular thing, there is a decent chance your child sorted it out a while ago too. Children are often clearer about this than we expect. They are used to things that talk and are not alive. They have had toys that talk their whole lives.
The half that should worry you
“The answer is true.”
And when I asked him straight out whether you can trust AI, he said yes. No hedge, no pause, no “sometimes.”
It took me a second to see why those two halves sit together, and then I could not unsee it. Knowing that nobody is behind it is part of why he trusts it.
Think about how a seven-year-old learns that information can be wrong. It is always attached to a person. Your brother says there is a shark in the pool. A kid at school says his dad has a helicopter. A grown-up says maybe, we will see. Every unreliable thing a child has ever been told came out of a human mouth, and people are unreliable for human reasons. They guess. They show off. They tell you what you want to hear. They are wrong and will not admit it.
Take the person away and what is left ought to be cleaner than that: no showing off, no motive, just the answer.
That is genuinely good reasoning. It is also exactly backwards, and it is the single most useful thing you can help your child turn around.
What to say next, and what not to
Do not say “it’s not always right.”
I know that is the instinct. It is also the least effective sentence available. Your child will nod, agree with you, and change nothing, because it is an abstract claim about a thing that has never once visibly failed in front of them. Agreement is not the same as belief.
What works is catching it in the act, together.
Sit down with them and ask it something it cannot possibly know. Not a trick question. Ask something from their own life: their teacher’s name, what they got for their birthday, how many people live in your house, the name of their best friend.
It will answer. That is the whole point. It nearly always answers.
Then you say the thing that actually does the work:
“It didn’t know that. It wasn’t lying to you. It just didn’t know, and it said something anyway.”
That sentence works because it is precise. A thing that cannot tell the difference between knowing and guessing will hand you both in the same voice. Your child does not have to distrust the machine to understand that, which is just as well, because they were never going to distrust it on your say-so.
Then let them do it again. Let them find the next one themselves. A child who has personally caught it being wrong three times has learned something a hundred warnings would not have taught them, and they got to be the clever one doing the catching, which is the only reason it will stick.
Answering the question he actually asked
The only question he had of his own was this: “How does AI work?”
Nobody had ever told him. He uses it, he trusts it completely, and he has no idea what it is doing. Most children in that position are in exactly the same spot, and so, if we are honest, are plenty of the adults handing them the tablet.
Here is the version that lands at this age:
“It read almost everything people have ever written. Now it is extremely good at guessing which word should come next. That is what it is doing the whole time: guessing the next word, very fast, over and over.”
And then the part that does the real work:
“It’s the best guesser anybody has ever built. But a guesser isn’t a knower.”
That framing beats “it’s a computer program,” which explains nothing and closes the conversation. It beats “it’s like a very clever robot,” which walks straight back into the problem. A seven-year-old understands guessing completely. They guess constantly. They know exactly what it feels like to guess confidently and be wrong, because it happened to them this week.
You are giving them the correct mental category, and the caution comes with it.
“How to save somebody”
One more thing from that conversation that I keep thinking about.
When I asked what AI is for, the first thing out of his mouth was emergencies. How to save somebody. How to stop something.
I do not know where that came from. Possibly something he watched. Possibly he had only ever seen an adult reach for it when something was actually wrong.
But it means that in his head, this is the thing you turn to when it matters most. Not for jokes, and not for stories, although he knew it could make those too. For the moment when somebody is hurt.
That is worth knowing about your own child, and you will only find out by asking. Ask them what they think it is for. The answer tells you what they will reach for it for when you are not in the room, which is the scenario all of this is about.
One child, and what to do about it
He is one seven-year-old. I am not going to pretend a single conversation is research.
But the shape of what he said is worth taking seriously, because it is not the shape we have been preparing for. We have been getting children ready to not fall in love with the machine. He was never going to. What he needed, and did not have, was any reason to think the machine could be wrong.
The two questions worth asking your own child this week:
- “Who’s talking, when it talks?” You may find they already have this one.
- “Does it ever get things wrong? How would you know?” This is the one to listen carefully to.
If the second answer is a shrug, or a no, you now know exactly where to start. And you know how to start: spend an afternoon catching it in the act together.