Mostly looking things up, homework and fun. In Pew Research Center's February 2026 survey of US 13 to 17 year olds, the top three uses were finding information (57%), schoolwork (54%) and entertainment (47%). Chatting for company came near the bottom. Those are averages, though, and averages tell you nothing about your child. The platform's age rules and your school's policy still come first.
Last updated October 21, 2026. I re-read the Pew, Ofcom and OpenAI pages behind this post today. What changed: the Ofcom figures now carry the age groups Ofcom measured them on (an earlier version put a 16 to 17 figure on all 13 to 17 year olds), the Pew and OpenAI claims now link to the pages they come from, and the section on ChatGPT's settings now follows OpenAI's current help pages.
What teenagers say they use AI for
The best picture we have comes from the teenagers themselves. Pew Research Center surveyed 1,458 US teens aged 13 to 17 and their parents between September 25 and October 9, 2025, and published the findings on February 24, 2026. Asked what they use AI chatbots for, teens said (Pew reports these as shares of US teens, with a margin of error of plus or minus 3.3 points):
- Looking up information: 57%
- Help with schoolwork: 54%
- Fun or entertainment: 47%
- Summarizing something: about 40%
- Making or editing images and video: about 40%
- Casual conversation: 16%
- Emotional support or advice: 12%
The British picture has the same shape, though Ofcom measures a wider age range, so the numbers are not directly comparable. Ofcom's Children and Parents: Media Use and Attitudes report, published on May 21, 2026, found that 56% of UK children aged 8 to 17 had used AI, rising to two thirds (66%) of 16 to 17 year olds. Among the 8 to 17 year olds who use AI, 51% said they use it for schoolwork or homework and 43% simply for fun. Ofcom notes that its comparison with earlier years (34% for schoolwork in 2023, 45% in 2024) is indicative only. One in ten AI users aged 8 to 17 said they had used it as someone to talk to.
Read those two sets of figures next to each other and the shape is the same in both countries. Most of what teenagers do with AI is ordinary: a search, a first draft, a joke, a picture of the dog as a Renaissance painting. That is not an argument for relaxing. It is an argument for aiming your attention at the right place. For how common AI is across all ages, including children under 13, how many kids use AI compares four recent US and UK surveys.
The ladder, not the switch
Here is the number that changed how I talk to parents about this. Among US teens in that Pew survey, 10% said they use chatbots for all or most of their schoolwork. Another 21% said some. Another 23% said a little. And 45% said they had not used them that way at all.
So "teenagers use AI for homework" is four different situations wearing one sentence. A child who has never opened ChatGPT for an assignment and a child who runs every assignment through it are both inside that headline, and they need opposite conversations. The first needs to be taught how to check an answer. The second needs their thinking handed back.
Teenage AI use is a ladder, not a switch. The useful question is not whether your child uses it. It is which rung they are standing on.
Parents are 13 points behind, and a third are not sure
In the same fieldwork, 64% of teens said they use AI chatbots. Only 51% of parents said their teen does, and 18% said their teen does not. That leaves about three in ten parents who said they were not sure, which is the figure I would sit with. Pew's page on parents' views also reports that just over half of parents talk with their teen about chatbots, and that roughly four in ten have not had these conversations.
My reading of it, and this is interpretation rather than something Pew measured: the gap is not teenagers hiding something. It is that AI stopped being a destination. Nobody announces they used the search bar. When it is inside the phone keyboard, the homework site and the app they already had, using it does not feel like an event worth mentioning, so it does not get mentioned.
The use you worry about is not the use they do most
Pew asked parents which uses they were comfortable with. About eight in ten were fine with their teen searching for information, roughly six in ten with schoolwork help, fewer than a third with casual conversation, and 18% with a chatbot giving emotional support or advice, the only use Pew asked about that a majority of parents were not comfortable with.
Now put that against what teens actually do. The two uses parents are least comfortable with, casual conversation and emotional support, are the two the fewest teens report. The uses almost every parent is fine with are the ones happening constantly. And if what worries you is the sheer amount rather than any one use, that is a different question with a different answer: how much AI use is too much, and what to look at instead of the hours.
I am not saying the worry is misplaced. If your teenager is leaning on a chatbot when they are low, that matters more than any homework question, and I have written separately about what to do when a teen talks to an AI about their feelings. If you are worried about your child's mental health right now, the route is a person, not a setting: your GP or pediatrician, the school's safeguarding lead or counselor, and in a crisis 988 in the US or Samaritans on 116 123 and Childline on 0800 1111 in the UK. Where your child's school has put an AI system into that pastoral role, it is worth knowing what those school AI counselors can and cannot do before you rely on one.
But the everyday risk sits somewhere quieter. Ofcom found that nearly eight in ten (78%) UK 8 to 17 year olds at least sometimes read the AI summaries at the top of search results, and that 46% of those who had read them agreed the information is always accurate. In the same research, just over half (52%) of 13 to 17 year olds who were aware of AI felt confident they could spot AI-generated content, and around six in ten (59%) of the 8 to 17 year olds shown an AI-generated image of a celebrity correctly identified it as AI. Those two figures come from different age groups, so they cannot be subtracted from each other, but they are a reminder that feeling sure and being right are different things.
That is the gap I would spend the year on. Not the confessional chatbot, which is real but rarer. The confident wrong answer, read at speed, on a Tuesday, and copied into a piece of work without a second thought.
What I See in Schools
When I train teachers, the thing they raise most about older students is not cheating. It is flattening. Work comes in that is competent, tidy, correctly structured and says nothing that the student could not have been told. The voice has gone out of it.
The pattern I see most often in secondary schools is a child who is not trying to cheat at all. They are trying to get to "finished" faster, on a night with three deadlines and a shift at work, and the machine offers a version of finished that looks convincing. Nobody has ever shown them what they gave up to get it.
Teachers spot this quickly, and usually not through detection software, which they trust less than the headlines suggest. They spot it because they know how that particular child writes.
The question that gets you a real answer
If you ask a teenager "do you use AI?", you will get a shrug and a yes, and you will learn nothing. Try this instead.
The Last Three Question: do not ask "do you use AI?" That gets a shrug. Ask "what were the last three things you asked it?" A general question gets a general answer. A specific one usually gets a phone held up.
It works because it is not an accusation and it is easy to answer honestly. The last three things are almost always mundane, which lowers the temperature, and the mundane answer is exactly the information you wanted.
Some sentences that work, said in the car or over the washing up rather than across a table:
"What were the last three things you asked it?"
"Show me one where it got something wrong."
"What does it do better than you, and what do you do better than it?"
"If you handed that in and your teacher asked you to explain the second paragraph, could you?"
That last one is the whole thing in a sentence. Let AI help with the doing; keep the thinking with your child.
What changes between 13 and 15, and 16 and 18
The band matters more than people expect, because the pressure changes.
At 13 to 15, the honest job is teaching the check. This is the age when the summary at the top of the search results becomes the answer, so make disagreement normal: ask the same question twice, in two tools, and look at where they differ. Keep AI use in shared rooms rather than behind a bedroom door, not because you are policing it but because a machine that is always agreeable is easier to lean on in private. Agree that first drafts can be theirs and edits can be shared, and be specific about which subjects that applies to.
At 16 to 18, monitoring stops working and the real risk becomes coursework, personal statements and applications. Two things are worth saying plainly. First, an admissions essay or personal statement written by a chatbot reads like one, and the people reading them have now read thousands. Second, exam boards and universities have their own rules about AI in assessed work, and those rules override anything agreed at home, so the sensible move is to read your child's school or college policy together rather than guess at it.
For younger siblings watching all this: under 13 is below the minimum age on most of these services anyway, and the guidance there is different. That is a separate conversation, and our guide for parents new to AI is the better place to start.
What the settings can and cannot do
Start with the rule, because it is the one thing that is not a matter of opinion. OpenAI's terms of use (the version it shows UK and international readers, effective January 1, 2026) say: "You must be at least 13 years old or the minimum age required in your country to consent to use the Services." They add that anyone under 18 must have a parent or legal guardian's permission. Other companies set their own ages, and those limits are self-declared, which means they stop an honest child and not a determined one.
OpenAI's help page on age prediction, which I read on October 21, 2026, says the system looks at signals linked to an account, including the topics discussed, the times of day it is used and how long the account has existed. If it predicts that someone may be under 18, ChatGPT automatically turns on the ChatGPT for Teens experience, with extra safeguards for sensitive content and some feature limits. Adults placed there by mistake can verify their age. The parental controls page, also read on October 21, 2026, says a linked parent can schedule quiet hours when their teen cannot use ChatGPT and switch features such as voice mode and image generation on or off. Study mode is off by default, so conversations only start in it if a parent turns that on.
Here is the limit, in OpenAI's own words: "Linking accounts does not give a parent or guardian access to a teen's conversations, chat history, or real-time activity." The same page says ChatGPT "may notify you in limited situations involving a serious concern about your teen's safety." That is a deliberate design choice and I think it is the right one, but it means no control you switch on tonight will tell you what your child is talking about. Only your child will. If you want the full walkthrough of what those settings do, we have a guide to ChatGPT's parental controls and one for Google Gemini.
Character and companion apps are a different category with different risks, and we look at those separately.
What to ask the school
Three questions, in this order, at parents' evening or by email to your child's head of year. They are short on purpose.
- Which AI tools are my child expected or allowed to use, and who decided? A good answer names tools and a person. A vague one tells you the policy has not reached the classroom.
- How does this subject expect AI to be used in coursework, and how is that checked? Rules differ between English and computer science, and your child needs the subject answer, not the school answer.
- What are you teaching students about checking what AI tells them? If the answer is nothing, that is worth knowing, and it is the gap you can fill at home this week.
What teachers are actually doing with AI is often more sensible than parents expect, and knowing it makes these conversations easier.
Where to start tonight
Ask the last three things. Do not correct what you hear. Then pick one of them and look at it together, and ask how you would check it if it mattered.
If your child's school has not yet spoken to parents about AI, that gap is worth naming, and it is the work I do with schools: training staff, setting a strategy and making the policy mean something in the classroom. It is worth sending them. I also write a newsletter that keeps teachers and parents up to date without the panic, and parents are very welcome.
Dan Fitzpatrick is a former secondary teacher and assistant headteacher who has trained more than 150,000 educators in over 30 countries. More about Dan.
