Usually not the conversation itself, but often enough about it to matter. If your child uses a personal ChatGPT account on their own device and their own wifi, the school sees nothing. If they are signed in with a school account, or typing on a school laptop, or sitting on the school network, the school may see the prompt, the site, or both. The device matters more than the app.
The short version, in four lines
Can the school read the conversation? Only if your child is signed in to a ChatGPT account the school owns and manages. On a personal account, OpenAI's own help pages say your child manages that workspace, not an administrator.
Can the school see it anyway? Often, yes, by a different route. A school laptop or Chromebook usually carries monitoring software that reads what is on the screen and what is typed, whichever site is open and whoever is signed in.
Does anyone actually sit and read it? No. Software flags a small number of moments it treats as a risk, and a person reads those.
The one thing to do tonight. Ask your child one question: when you use AI for schoolwork, are you on your own account or your school one, and whose laptop are you on? That answer tells you almost everything.
Why there is no single answer to give your child
Three separate things decide what a school can see, and almost everybody collapses them into one. I call it the Sign-In Check.
The Sign-In Check is the question to ask before you ask what the school can see: what did your child sign in with, what were they typing on, and whose wifi were they using? The account decides whether anyone can read the conversation. The device decides whether anyone can read the screen. The network decides whether anyone can see which site was open. Change any one of the three and the answer changes.
The account decides the conversation
Schools and colleges can buy a managed version of ChatGPT. Where they have, the administrator's access is real and it is broad. OpenAI's help center, read on September 21, 2026, lists what an administrator of a managed account may access: "Content you submit to ChatGPT (prompts, files you upload, and outputs you generate)", "Conversation history and shared workspace content", and sign-in and device metadata. Its separate compliance documentation confirms that conversation messages are among the categories an organization can pull out, and that a workspace owner has to grant that particular permission deliberately rather than it being on for everyone.
Now the part parents miss, and it is the good news. The same OpenAI page states the limit plainly: "Unless your personal workspace has been verified or claimed by a ChatGPT Enterprise or Edu workspace, you manage your personal workspace, not an admin." A personal account your child opened with a personal email is not readable by their school. Not by the principal, not by the IT manager, not by the safeguarding lead.
So the question is not "does my child use ChatGPT". It is "which ChatGPT is my child signed in to".
The device decides the screen
This is the layer that catches people out, and it undoes the reassurance above.
A school-issued laptop or Chromebook is not a neutral object. In England, the Department for Education's filtering and monitoring standard, last updated on September 16, 2026, expects schools to run monitoring that includes "device monitoring using device management software" alongside network monitoring, to produce "weekly monitoring reports highlighting incidents", and to send "immediate reports when an incident is classed as high-risk". In the United States, schools taking E-Rate discounts have to certify an internet safety policy, and the Universal Service Administrative Company's guidance, dated March 5, 2026, puts it in one line: "For schools, the policy must also include monitoring the online activities of minors."
Monitoring software of that kind does not care which website is open. It reads text as it appears. So a prompt typed into ChatGPT on a school laptop can be flagged exactly as an email or a document would be, even where the child is signed in to a personal account, and even where the school has never bought an AI product in its life.
That is not theoretical. Euronews reported on October 8, 2025 that a 13-year-old in Florida was arrested after a violent question typed into ChatGPT on a school laptop triggered an alert from the school's monitoring system, which went to the officer assigned to the campus. The child said he was joking. I mention it for one reason, and it is not about that family: it shows the route. The school did not read his chat history. A piece of software read one line as he typed it.
The network decides the address
The thinnest layer, and the one people assume is the strongest. On school wifi, traffic logs generally record which sites a device reached and when. That is an address, not a conversation. Knowing that a laptop opened chatgpt.com at 11:14 tells a school nothing about what was asked.
Worth knowing on the Google side, because so many schools run on it: a Workspace administrator can search a long list of log events, which as of September 2026 includes Chrome, Devices, Drive, Classroom and Gemini for Workspace activity. Those logs are about actions and access. Google's own Chrome Enterprise and Education documentation on managed profile reporting describes what it returns as "profile information, browser version, installed extensions, applied policies, and more", and does not list browsing history or page content among them. The URL-level and content-level watching, where it happens, comes from the filtering system or the monitoring product a school has bought, not from the admin console by itself.
What I See in Schools
I spent years as a Director of Digital Strategy in further education, and I work now with schools, trusts and districts on their AI strategy. The pattern is consistent, and it is not the pattern parents expect.
Nobody is reading your child. There is no adult with a screen of transcripts. What exists is a queue of alerts, most of them nonsense, generated by software that cannot tell a history essay about the Civil War from a threat, and a designated safeguarding lead who works through that queue alongside everything else in the job. When a school does look at a child's AI use, it is almost always because an alert surfaced, or because a teacher suspected a piece of work was not the child's and went looking at document history instead.
The gap that matters is this, and it is why I would rather a parent asked than assumed. Most schools I work with can tell me which filtering and monitoring product they run. Far fewer can tell me what it does with AI prompts specifically, who reads the alerts at four o'clock on a Friday, or what happens between an alert being raised and a parent being called. Those are answerable questions. They are just not usually asked until something has already happened.
A school rarely sees what your child asked an AI. It sees the handful of moments a piece of software decided an adult should look at, and it sees them on a school device, whoever your child was signed in as.
What this looks like by age
Under 8. Largely academic, because a child this age should not have an AI chatbot account of their own. If they use AI at school it is inside a teacher-controlled tool, and the teacher sees the output by design.
8 to 12. This is where school devices go home. Set the expectation now, before there is anything to hide: the school laptop is the school's, and someone can see what happens on it. Treat that as ordinary rather than as a warning.
13 to 15. The age of the personal account and the strong feelings about privacy. The honest line is that a personal account on a personal phone is genuinely theirs, and a school laptop genuinely is not. Children who understand which is which make better decisions than children who assume everything is watched, because assuming everything is watched is what pushes them onto the device you cannot see at all.
16 to 18. Bring in the thing that outlasts school. Anything typed into an account an institution owns can be read by that institution, and that will be true of their first employer too. That is not a rule you impose. It is a fact about the adult world that they are about to enter.
You could say
For a younger child, on the school laptop:
"That laptop belongs to school, so school can see what happens on it. That is not because anyone thinks you would do something wrong. It is the same reason they can see your exercise book."
For a teenager who has just discovered the question:
"Honestly? On your own account, on your own phone, no. On the school laptop, yes, they can see what you type, whichever site you are on. So it is worth knowing which one you are on."
And if you suspect the real question underneath is about getting caught using AI on homework, that is a different conversation and a better one. What a teacher actually notices, and what to do when a child is accused, is covered in what to do when your child is accused of using AI.
What to ask the school
Four questions, and any school should be able to answer them:
- What filtering and monitoring do you run, and does it cover text typed into AI tools as well as websites visited?
- Who reads the alerts, and what happens out of hours?
- At what point are parents told, and are we told when an alert turns out to be nothing?
- If our child uses AI at school, is it through a school-managed account, and if so who can see those conversations?
The fourth one is the question of the moment, and it is where good schools and drifting ones separate. If you want the fuller picture of how AI use, monitoring and policy should hang together, the way I would check whether a school has thought this through is set out in my school AI readiness work. It is a reasonable thing to send to a head of year.
What none of this can guarantee
Monitoring does not make a child safe. It produces alerts, and alerts are only as useful as the adult who reads them and the relationship that follows. The Center for Democracy and Technology, whose nationally representative US surveys published on October 8, 2025 covered 806 teachers, 1,018 parents and 1,030 high school students about the 2024 to 2025 school year, found that 31 percent of students said they had held back-and-forth conversations with AI for personal reasons on a device provided by their school, and that 38 percent agreed it was easier to talk to AI than to their parents. My reading of those two numbers together is that the personal conversations are happening on school hardware, and that a monitoring queue is a poor substitute for the second finding being less true.
Nor does a personal account mean nobody sees anything. It means the school does not. The company still holds the conversation, which is a separate question, and one I have answered separately in what parents can and cannot see of a child's ChatGPT history. If you want to change what is kept, the settings that exist are in the ChatGPT parental controls guide.
And where the school's own rule is stricter than anything here, the school's rule wins. Some schools ban AI on their devices outright, and a child who works around the ban has a discipline problem rather than a privacy one. If a monitoring alert leads to a safeguarding conversation, the school's designated safeguarding lead is the person to speak to, and their judgment on a child's wellbeing comes before anything a parent reads online, this page included.
Let AI help with the doing; keep the thinking with your child. Knowing which account they are signed in to is part of the doing, and it takes one question at the kitchen table.
If your child's school has not yet spoken to parents about what it monitors and who reads it, that conversation is overdue, and it is the kind of thing I help schools get right. I also 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, assistant headteacher and Director of Digital Strategy who now helps schools and families make sense of AI. More about Dan.
