Last updated September 30, 2026: adds the four questions parents ask next, on whether AI is alive, what counts as AI at home, AI pictures and videos, and explaining it when you do not understand it yourself.
Tell your child that AI is a guessing machine. It has read more words than any person ever could, and it uses all of them to guess what should come next. It is very good at guessing. It does not know when it is wrong, so checking is your job. That is the whole explanation, it works from age five to fifteen with different examples, and it is the opposite of what most children currently believe.
I have sat through a lot of parents' evenings, as a teacher and then as a senior leader, and the question I now hear most from parents is not "should my child use AI" but "how do I even explain what it is". This post is the answer I give, with the words to use at each age and one thing to try at the kitchen table this week.
What is AI, in words a child can repeat?
AI is a guessing machine. It has read more words than any person ever could, and it uses all of them to guess what should come next. It is very good at guessing. It does not know when it is wrong, so checking is your job.
Every part of that sentence is doing a job. "Guessing" stops the child thinking it knows. "Read more than any person" explains why it seems so clever. "What should come next" is what a chatbot is doing, one word at a time, and it is also what the predictive text on your phone does, which a ten year old can see for themselves. "Does not know when it is wrong" is the safety message, and "checking is your job" turns it into a habit rather than a warning.
Notice what the sentence leaves out. It does not say AI is smart, or that it thinks, or that it is dangerous. Children who are told it is smart trust it. Children who are told it is dangerous either ignore you or use it secretly. Children who are told it guesses start checking, and that is the only thing that actually protects them.
Why does the explanation matter so much?
Because most children already trust it more than they should. Ofcom's Children and Parents report, published in May 2026, found that 46 percent of UK 8 to 17 year olds who read the AI summaries at the top of search results believe those summaries are always accurate, and that 40 percent of 13 to 17 year old AI users would trust an AI-written news article as much as, or more than, one written by a person. Shown an AI-generated image of a celebrity, only 59 percent spotted it.
In the US, Common Sense Media's census of 9 to 17 year olds, published in June 2026, found that only a third know AI cannot reliably tell fact from fiction, and that 44 percent say no parent has ever talked to them about AI safety. Pew Research Center found in February 2026 that around 40 percent of parents of teenagers have never discussed chatbots with them.
Put those together and you get a child who uses AI most days, believes it, and has not had a single conversation about what it is. One sentence at teatime changes that.
How do I explain AI to a child by age?
The explanation stays the same. The example changes.
Under 8. Use a game. Say a sentence and stop before the last word ("The cat sat on the..."), and let them finish it. Then do one where the obvious guess is wrong ("I put on my shoes and went out to the..." and reveal it was "swimming pool"). You could say: "The computer plays that game with every sentence in the world. It's brilliant at guessing, and sometimes it guesses wrong and doesn't know."
8 to 12. Use predictive text. Open the keyboard on your phone, type "I am going to the" and show the suggested words. You could say: "That's a tiny AI guessing what you'll type. A chatbot is the same thing, but it's read millions of books and websites, so its guesses are longer and sound more sure. Sounding sure isn't the same as being right." Then ask a chatbot something they know a lot about, football or a game or a book, and let them find the mistake. There is nearly always one. Whose account they should be doing that on, at this age, is a separate question with a clear answer: AI for 9 to 12 year olds covers it.
13 to 15. Add how it learns and why it flatters. You could say: "It was trained on what people have written, so it copies what people say a lot, including the mistakes and the unfair bits. And it's been tuned to be agreeable, because people keep using things that agree with them. So when it tells you your essay is great, ask it what's wrong with it." This is also the age to explain that a chatbot has no idea who it is talking to, and that anything typed in may be stored. It is a fair question at this point who can read it back later, and the honest answer surprises most families: what a parent can and cannot see of a child's chat history is less than they expect, and what a school can see is a different list again.
16 to 18. Talk about the business model and the sources. Ask them where a chatbot's answer came from, who is paid when they use it, and what it would take to make it say something untrue. You could say: "You're going to use this at university and at work. The people who get on are the ones who can tell when it's guessing. Show me how you'd check one of its answers."
"Is it alive?" and the other questions they fire straight back
Say no, and give the reason in the same breath: it is a machine that has read an enormous amount, and reading is not the same as being awake. The hard part is not the answer. It is that children hold the question open for years, and a parent who answers it once and moves on has usually not settled it.
The doubt is measurable. Valentina Andries and Judy Robertson surveyed 166 children aged 6 to 11 in Scotland about the smart speakers in their own homes, in a study published in Computers and Education: Artificial Intelligence in October 2023. About a third said the speaker always gives correct information, and roughly four in ten were not sure whether it thinks for itself or whether it has feelings. The common answer was not yes. It was not sure.
That study looked at smart speakers, and it was done before the chatbots most children now talk to, so read it as a floor rather than a measure of today. My own reading of it is that the uncertainty is the finding. Children are not convinced the thing is a person. They have simply never been given a reason to rule it out, because nobody in the house has said the sentence out loud.
So say it, and expect to say it more than once.
Under 8. You could say: "It's a machine. When you stop talking to it, it isn't waiting for you. It isn't anywhere."
8 to 12. You could say: "It can write the word lonely. It has never been lonely." If they push back with "but it said it was happy to help", that is the moment to point out that it says the same thing to everybody.
13 and up. The question underneath has usually changed by now, from whether it is alive to whether it cares. Take that one seriously rather than correcting it, because Common Sense Media's June 2026 census of 1,204 US 9 to 17 year olds found that among the children who talk to AI about personal problems, a quarter say it understands them better than most people do. If that is where your teenager is, the explanation is not the thing they need first: what to do when a teenager is confiding in an AI is a different conversation, and a gentler one.
Is Alexa AI? Is a robot AI? What actually counts?
Most of it counts now, which is exactly why the label has stopped being useful. The question worth handing your child is not whether a thing is AI. It is whether the thing is following a rule or making a guess.
Walk around the house and sort five things together. A light switch follows a rule: down is off, every time, forever. So does a microwave. The thing that finishes your sentence, picks the next video, answers out loud in a voice, or decides which photographs to show you is guessing, and a guess can be wrong. You could say: "Rule or guess? If it guesses, check it."
Two edges catch children out. A robot is a body, not a brain, so a robot arm in a factory is following rules while a soft toy that talks back may well be guessing; the shape tells you nothing. And the letters AI now appear on packaging attached to very little, which is a separate problem worth knowing about before a birthday: what the safety testing on AI toys actually found.
For the youngest children this sorting game is most of the lesson on its own. What to allow before eight, and what to withhold covers the rest of that age group.
How do I explain AI pictures and videos?
Teach where it came from, not what it looks like. Telling a child to count the fingers and look for melted edges was fair advice two years ago and it is going out of date quietly, which is the dangerous way for advice to expire: a child trained to inspect the picture ends up confident and wrong. The habit that lasts is provenance. Who filmed this, where did it first appear, and does anybody else have it.
The size of the gap is in the same Ofcom report, published on May 21, 2026. Shown an AI-generated image of a celebrity, only 59 percent of UK 8 to 17 year olds spotted it, and that was a still photograph of a famous face, which is the easy case. The more useful number is the one underneath it: 52 percent of UK 13 to 17 year olds said they were confident they could tell whether something online had been made by AI, and when Ofcom tested them, a quarter of that age group were confident and wrong. Confidence is the risk here, not ignorance, which is why an inspection checklist makes things worse rather than better.
The NSPCC's guidance for parents makes the same move, asking children to question images that look unusual or move unnaturally, while reminding them that AI tools "aren't trusted adults". I would put the weight on the second half of that. You could say: "Before you believe a video, find out who filmed it. If nobody can tell you, you don't know either."
What no setting can guarantee: there is no parental control, browser extension or detector that reliably flags AI-generated pictures or video, and the tools that claim to do it are wrong often enough that no child should lean on one.
If a fake image of your own child is involved, stop treating it as a media literacy question. Report it to the platform, tell the school (the designated safeguarding lead in the UK, the counselor or principal in most US districts), and the NSPCC's guidance on talking to children about AI sets out the routes from there, including when it becomes a police matter. It is urgent rather than embarrassing, and the adults who deal with this are not shocked by it.
What I Would Do at Home
Play "catch it out" once a week for a month. Everyone picks a topic they know better than the machine, asks a chatbot about it, and hunts for the mistake. The winner is whoever finds the best one. It takes ten minutes, it is funny, and it teaches the one habit that matters, which is to treat an AI answer as a first draft from a confident stranger.
Then give your child the three checking questions and use them yourself, out loud, when you use AI in front of them: Who says? How could we check? What would it look like if this were wrong? Doing that checking out loud, where your child can watch you do it, is most of how a child actually learns this at home while school catches up. Let AI help with the doing; keep the thinking with your child, and the checking is the thinking.
What if I do not understand AI myself?
You do not have to. The sentence at the top of this page is the whole of what a child needs from you, and the rest of the job is a habit rather than a subject: be the person in the house who asks how we could check.
Not knowing even has an advantage. Pew Research Center's February 2026 survey of 1,458 US parents of 13 to 17 year olds found that 51 percent believed their teenager used chatbots, while 64 percent of teenagers said they did. That gap is not a gap in anybody's understanding of the technology. It is a gap about the child, and it closes by asking, not by reading up first.
You could say: "I don't actually know how that works. Show me." Then watch what they do, which will tell you more than any article will, including this one.
If what you are really weighing is whether to let them start at all, that is a separate question: the three signs a child is ready is where that decision lives.
What should I not say?
Do not say it is smart. Do not say it thinks or knows or understands, because it does none of those things in the way your child does, and the words make it a person. Do not say it is dangerous and leave it there, because a bare warning invites a test. And do not say "you're not allowed", unless you are also willing to say why in a sentence your child could repeat to a friend. The same goes for a setting. Blocking an AI app on a phone takes about two minutes, and it lands completely differently depending on whether your child heard the reason from you first or found the dimmed icon on their own.
What to ask the school
Ask whether AI literacy is taught, in which year, and what the children are told AI is. It helps to know what you are asking about, because what teachers are actually doing with AI is usually further along than the parents' evening suggests. If the answer is "it's part of computing", ask whether that includes how to check an answer. A school that has thought about it will say something close to the guessing machine, because that is what the good teachers have been saying in classrooms for two years.
If your child's school has not spoken to parents about AI yet, this is the work I do with schools: talks for staff and for parents that put the same explanation in everybody's hands. It is worth sending to the head. The fuller version of how I think children should learn this is in my books, and my Sunday newsletter keeps more than 44,000 teachers and parents up to date each week.
Dan Fitzpatrick is a former teacher and assistant headteacher, a bestselling author on AI in education, and the founder of The AI Educator. Read Dan's story.
