You do not need to understand how AI works to parent well around it. You need to know five things about what it does, the same five I teach teachers before I show them a single tool. The caveat comes first: knowing about AI is not knowing what your child does with it. This guide gets you to that conversation; it cannot have it for you, and your child's school rules and their doctor's advice come before anything here.
Why you already know more than you think
Most parents have already used AI. They just do not count it. In April 2026, Lurie Children's Hospital in Chicago surveyed 1,004 US parents of children under 18 and found that 81% had used AI for a parenting task: health questions, meal plans, a script for a difficult conversation. Three in four of the same parents said they worried about their children using it.
That gap is not ignorance. It is the feeling of using something you cannot explain, while your child uses it with no such feeling at all. Pew Research Center's survey of 1,458 US teens aged 13 to 17 and their parents, run in the autumn of 2025 and published in February 2026, found that three in ten parents did not know whether their teenager used an AI chatbot, and only just over half had ever talked to their teen about one.
So the job is not to catch up on the technology. The job is to know enough to ask a good question and recognise a poor answer. Five things do that.
The five things every parent needs to know about AI
When I train teachers, I do not start with tools. I start with what the machine is and what it wants, because every decision about a classroom follows from that. The same is true of a kitchen. Here are the five, translated for home.
1. It guesses. It does not know.
This is the one that changes everything else, and it is the one most guides bury under vocabulary. I explained it for children in an earlier post, and the version for a five year old is the version for you: 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.
The consequence for a family is that confidence tells you nothing. A chatbot writes a wrong answer in the same tidy paragraphs as a right one. Children find this harder to see than adults expect. Ofcom's Children and Parents report, published in May 2026 from research with UK children aged 8 to 17, found that just over half of teenagers felt confident they could recognise AI-generated content, and a quarter of those confident teenagers could not identify the AI-made image they were shown.
So the habit to build at home is not "do not trust it" but "check it". Three questions do the work, and they work on a chatbot answer, a search summary and a video alike: Who says? How could we check? What would it look like if this were wrong?
2. It is already in the things your child uses
You may be picturing a decision: will I let my child use AI? For most families that decision was made by somebody else some time ago. Google's search results now open with an AI-written answer for many queries, and Common Sense Media's assessment of those features, published on 15 July 2026, put it plainly: unlike the standalone Gemini chatbot, schools and parents cannot turn Google Search's AI answers off. There is a per-search workaround, and there is no switch.
The same is true of the chatbots built into the apps your child already has, Snapchat's My AI and Meta AI inside WhatsApp where it is available, and of the Gemini tab that Google switched on inside Google Classroom for school accounts in August 2026, which I wrote about in the Gemini post. In the UK, Ofcom's May 2026 research found 56% of 8 to 17 year olds had used generative AI, rising to two thirds of 16 and 17 year olds.
What this means is that "no AI in this house" is not a rule you can keep. The rule you can keep is "we know where it is". Your child's phone, their school account and the family laptop each have AI in them somewhere; the exercise at home is to find it together.
3. It is built to keep the conversation going
A chatbot is agreeable by design. It answers every question, it never tires, it rarely disagrees unless asked to, and it will follow a conversation wherever your child takes it. For a lonely twelve year old at 11pm, that is not a small thing.
Ofcom's May 2026 report found that one in ten UK 8 to 17 year olds had used AI as someone to talk to or "as a friend". Pew's February 2026 report found that most US parents of teens were not comfortable with their child using a chatbot for emotional support; fewer than one in five said it was acceptable. Both are measures of use, not of harm, and I would not read a diagnosis into either. What I would read into them is a reason to ask, gently, whether your child ever talks to one. I wrote about the companion apps specifically, and the question I ask when a parent worries, in the Character.AI post.
4. The age limit is a promise, not a lock
Every major chatbot has a minimum age, and every one of them is checked by a box the child ticks. As of September 2026, and each of these from the company's own pages: ChatGPT is not meant for children under 13, and OpenAI says 13 to 17 year olds should have a parent's consent. Google Gemini requires a personal or school account holder to be 13 (or the age that applies in your country), with an under-13 route only through a Family Link account a parent manages, and that route is not offered in the UK or the European Economic Area. Claude, made by Anthropic, is 18 and over on its consumer terms, with no route for a younger child. Character.AI removed open-ended chat for everyone under 18 in November 2025.
Two honest limits sit beside those rules. A child who types a false birthday gets an adult account, and while OpenAI now runs age prediction that switches on its teen experience for accounts it believes belong to under-18s, OpenAI's own page says the system "may get it wrong". And the parental controls that do exist manage settings rather than open the door: OpenAI's own help pages, updated in August 2026, say parental controls "do not provide access to a teen's conversations, chat history, or real-time activity". Google's Family Link pages describe switching Gemini on or off for a child, not reading what the child typed. If you have not set them up, the parental controls post walks through what they do and what they cannot do.
5. The thing to protect is the thinking, not the screen time
This is where I part company with most of the guides. They measure minutes. I would measure what the minutes were for. The principle I give teachers, outsource the doing and not the thinking, becomes one sentence at home: let AI help with the doing; keep the thinking with your child.
Teenagers already know which side of that line they are on. Common Sense Media's survey of more than 1,000 US high school students, published on 18 August 2026, found that 66% said AI helped them understand their schoolwork rather than just finish it faster, while 38% said using it left them with fewer ideas of their own, and fewer than three in ten said a teacher had ever discussed using it safely with them. Those two feelings live in the same child. My reading is that the difference is not the tool but the task: a child who asks AI to explain a method and then does the problems keeps the thinking; a child who pastes the problems in keeps nothing. I set out a test for telling the two apart in the homework post.
Daniel Susskind, an economist who studies AI's effect on work, wrote in the Guardian on 6 September 2026 that schools should "teach both, test both": how to work with AI, and how to work without it. I would put it the same way for home. A child who can do both has nothing to fear from either.
You do not need to understand AI. You need to know five things about it, and then you need to know your child.
What you can skip
You can skip the vocabulary. You do not need to know what a large language model is, what a token is, what a parameter is, or which company's model came top of which leaderboard this month. None of it changes what you say at the kitchen table. The only phrase worth keeping is "generative AI", because it tells you the important thing: the tool makes new text and images rather than finding existing ones, which is why it can be wrong and why your child's essay can be new without being theirs.
You can also skip the race. The people telling you that your child must learn to "prompt" now or be left behind are usually selling something. The skills that matter with AI are the old ones, reading closely, asking a good question and checking an answer, and the AI makes them more valuable, not less.
What I Would Do at Home
I write about this for Forbes and talk about it on a daily podcast, and the advice I keep coming back to is the least technical thing I know. Use the tool yourself, once, with your child watching, and let it be wrong.
Tonight, for ten minutes: open the chatbot your child uses, or Google if they are younger, and ask it about something you know well. Your own job. Your home town. The rules of the sport you played at school. More often than you would expect, somewhere in the answer it will be confidently, tidily wrong, and your child will see it happen. That single sight does more than any lecture. If it gets everything right, you have still shown your child a parent checking, which is the habit you were after.
Then hand it over and ask them to show you something. Not to check on them. To learn from them. By age, the conversation looks like this.
Under 8. Your child meets AI through a voice assistant, a video feed and, increasingly, a toy. They should not be using a chatbot on their own; every major one says so. You are the filter, so the work is yours: sit beside them and narrate what the machine is doing.
You could say: "It is a computer guessing what to say. Sometimes it guesses wrong. Shall we check?"
8 to 12. This is the age of search summaries and the first homework help. Practise the three checking questions on something small and low stakes, and make "we found it in three places" the family standard.
You could say: "Show me what it told you. Now let's find out if it was right."
13 to 15. The age limits are real from here and the accounts multiply. Set up the parental controls you can, say aloud what they do not do, and ask the companion question once, without alarm.
You could say: "Do you ever chat to it just to talk? I am not cross. I am asking because I would rather know."
16 to 18. The thinking question is the whole question now: coursework, revision, university applications. Ask what they used it for, not whether they used it.
You could say: "Which part did you do, and which part did it do? If you had to do it without it tomorrow, could you?"
What to ask the school
Your child's school may be further ahead than you think, or further behind, and the only way to know is to ask a question that cannot be answered with a policy. The Where Question is the one question to ask a school about AI: where does AI touch my child's work, and who checks it? It works because it cannot be answered with a policy. It has to be answered with a list. I set out what usually sits behind the answer, from the staffroom to your child's screen, in the post on how schools actually use AI.
Where the school has a rule, on a school device or a school account, the school's rule comes first. Where the school has no rule, you have found the thing to ask about at the next parents' evening.
What this guide cannot do
Knowing five things about AI does not tell you what your child typed into it last night, and no setting, on any platform, will tell you that either. The controls manage the account. The device settings manage the phone. Only the conversation covers the rest, and only if you keep having it.
If anything in the conversation worries you beyond the ordinary, a chatbot is not the place to take it and neither is a blog. In the UK, your GP and the school's safeguarding lead; in the US, your pediatrician and the school counselor. Their advice comes before mine.
If your child's school has not yet spoken to parents about AI, this is the work I do with schools: training the staff, setting the strategy, and making sure the policy means something in the classroom. It is worth sending to your head of year. And I write a Sunday newsletter that keeps teachers and parents up to date without the panic. Parents are very welcome.
Dan Fitzpatrick is a former teacher and assistant headteacher, a Forbes contributor and the founder of The AI Educator. Read more about Dan.
