Technology, Connected - How Do Kids Use ChatGPT In School?

Episode Date: October 1, 2025

A quarter of British 8 to 12 year olds have used generative AI. ChatGPT is the most popular tool, Snapchat AI is second. Half of teachers say they've already received homework that was clearly written by AI. And 58 percent of private school kids have used AI compared to just 18 percent at state schools, which is a gap so large it's reshaping the next generation's relationship with this technology before any of us have figured out what we want it to be. On this episode of Thinking on Paper, we unpack the Alan Turing Institute's new report on the impacts of generative AI on children, the largest study of its kind to date, based on a survey of 780 children, 1,001 teachers, and a workshop with 9 to 11 year olds in Scottish state schools. The report was funded in part by Lego, which is a small detail with a lot of meaning when the people who built the world's best analog play company decide to fund research on what the next thing is doing to kids. Along the way: why teachers are worried about critical thinking and parents are mostly positive, why ChatGPT is being used to "create fun pictures, find information, and entertainment" in that order, what the kids themselves said in their own words about misinformation and AI safety, the eight recommendations the Institute made for parents, teachers, and policymakers, and the one recommendation we found genuinely surprising — that kids themselves should be involved in designing the AI tools they're going to grow up using.--TIMESTAMPS(00:00) Disruptors And Curious Minds(02:40) The Alan Turing Research On The Impact Of AI On Children(03:49) The Most Popular AI Platform For Kids(04:41) How Are Kids Using AI?(04:55) Key Statistics From The Research On AI Use In School(05:33) The Divide: Private vs. Public School AI Usage(08:08) Are Kids Using AI To Cheat In School?(09:55) The RITECH Framework: Evaluating AI for Kids(11:18) Quotes From Kids On AI(16:13) Child Versions Of ChatGPT(18:12) Recommendations From The Alan Turing Institute--Other ways to connect with us:⁠Listen to every podcast⁠Follow us on ⁠Instagram⁠Follow us on ⁠X⁠Follow Mark on ⁠LinkedIn⁠Follow Jeremy on ⁠LinkedIn⁠Read our ⁠Substack⁠Email: [e-mail]

Disruptors and curious minds, parents, parents, I'm talking to you. If you have children between the ages of 1 and 18, if you're thinking of having children, if you know somebody with children, listen up. This is for you. We're talking about the impact of generative AI on kids, how they've been used in schools, how the teachers are using AI, how the teachers and the students together are using AI. We're reading the Alan Turing's Institute on the impacts of generative AI on children. There was the biggest questionnaire, I think, that's ever been done on children's use of AI.

There was a three-day workshop where kids were exposed to AI. They used AI in conjunction with their teachers to get to the root of their feelings, their opinions. There's some very insightful recommendations for parents, for teachers, for policy makers, and some very interesting insights into how children are using AI today, Jeremy. Where do you stand on all this? What do you feel about the use of AI by kids in schools? before we get into the paper. I've got four kids.

Mark, you have two kids. Mine go all the way up from in college down to six grade. Mark, yours are how old? My kids are nine and six, so they fall right into the bull's eye of this paper that we're going to break down. Totally. Yeah. So we're coming from the parent side as well on this. I started thinking about this study that the closest platform or the closest technological

comparison to AI. Not saying it's the same thing, but it's a self. similar thing that came out that affects a generation a certain way. I think about social media and what that's doing or what that is done to kids. AI, generative AI in particular can be more potentially be more helpful than social media, but also can on the flip side be, man, maybe even more dangerous. The upside is astronomical. There just has to be a framework.

There has to be a structure of how kids and adults, Today we're talking about to use it and are exposed to it and behave with it. Before you get going, I just want to leave a couple of questions for our listeners as we're talking through this paper, as you're thinking about what we're speaking and thinking about. What is a healthy AI relationship between children and AI? My daughter is nine. What should she be doing with AI in her school, if anything? And with that in mind, let's get going.

So the paper is based into two different reports. One is a survey of 780 children between the ages of 8 and 12 and their parents. And 1001, who is that lucky one? 1,0001 teachers with kids in primary school or secondary school. And I point out that this survey was in the UK. Not teachers with kids in primary or second, but teachers teaching kids in primary. Teachers, teachers, yes.

Good, important to notice that different. And the second report was with 40 children between the ages of 9 and 11 in two state-funded schools in Scotland. And we'll get into that after the first one. And that was a three-day workshop, less kids. But that was more about the feelings, the emotions, the children's opinions on A. And there's some very insightful, very interesting, very powerful quotes from some of the kids, which we'll read later. But yeah, first, the survey. Yeah, the survey.

And again, this was a study through the Allen-Turring Institute was funded or partially funded. by Lego, which is pretty interesting. The OGs of analog play, the OGs of IRL play and creativity. I thought that was cool that they're supporting this. I don't know if this is unfortunate news or not, but just given what we're learning about chat GPT and OpenAI, the number one tool being used by kids is actually chat GPT. How do you feel about that one? I'm not surprised.

What I was surprised, which is because of my age, that one of the second, I think it was Snapchat AI was up there. Snapchat, I think. Yeah, yeah. Snapchat AI. So, yeah, I wasn't surprised by ChatGPT. Snapchat's pretty big. I've resisted it, but my wife actually got it and has this nice communication channel

with our kids. Not saying it's only through Snapchat. We go out of our way to generate or try to orchestrate real life moments and stuff. But she now like snaps them. So, I mean, it's a big part of their world. So we know ChatGPT. is the most popular tool. How are they using it?

What did the study find how they're using it? The three main use cases were, quote, creating fun pictures, finding information and entertainment in that order. I've got a few key results from the paper. I'm just going to firebom them out there just so we can know. 55% of households in the UK use AI. 25% of 8 to 12 year olds have used AI. 58% of 8 to 12 year olds in private school have said that they've used AI,

generative AI, only 18% in state schools. There's a lot to read into that, which we'll get to. 76% of parents are positive, 72% of teachers worry about AI, 50% of teachers report students submitting work done by AI, and two-thirds of teachers use AI themselves. So there's a lot going on. It's there. It's very prevalent. I don't know much about the gap between private schools,

and public schools in the UK, is it different than in the States? That statistic that 58% of private school kids between ages of 8 and 12 have said they've used generative AI, but only 18% at state school, I think that statistic gives you a very good window in the difference between state and private schools in the UK. It's big. And if 58% of them are using AI already and only 18% are, the have and the have not, which we speak about a lot on thinking on paper, not just in AI, about all technologies, about this balance of power is already there.

So some of the recommendations that they get, we'll get into later, involved trying to close that gap. Yeah, it just accelerates, like technology does, it accelerates the, what's already happening from an analog perspective. I've got a question for you then, as an American, I'm sure it's probably the same in other countries. Why are kids in private schools, what, 20, three times, more likely at the age of 12 to have used AI? I don't know.

I don't know what to think about that. Oh, come on. Yeah, it's the parents. It's social and economic backgrounds and the parents. 10-year-olds aren't using AI unless their parents are, A, using it and B, allowing them to use it. So the parents at these schools are using AI more and that they're letting their kids use it. Yeah, it's familiarity. I guess, you know, if your parents talk about XYZ, you're going to be more familiar with

XYZ. If they have a good opinion about XYZ, you're a little. likely going to have a good opinion about XYZ until you could form your own opinion through experiences, I guess. That's interesting. Do you think wealthy people care more about speaking about artificial intelligence and technology in general than people from less wealthy economic backgrounds? Just to put you on the spot, Jeremy. Yeah. I don't know. I wouldn't say I'm in the upper 1% of wealth, the wealth echelon by any stretch of the imagination. Yeah, because if you were, So that's something you're not telling me.

But yeah. No, I'm definitely not. But like, I'm very deep into my awareness of technology and as you are as well. So, I mean, there are people like us. They're curious individuals. There are maybe, I don't know, people with better access to understanding what the hell these things are. Maybe they work for companies that are already implementing some of those things.

If your company pays for your chat GPT subscription or you're using your personal account, let us know. We would like to know what people are you. And why? And why? Yeah. Let's talk a bit more about some of the other things we think about with AI. Are kids using it to cheat on their homework?

50% of teachers have reported that, that they've seen a homework submitted by AI. Is that 50% over exaggerated or under exaggerated? Is there some slipping through the net? I think there's some slipping through the net. Probably so, but 85% of teachers are showing that it increases their productivity on what they're doing. So it's about becoming more efficient at what you're doing. But those same teachers were also worried about some of the impacts that it could have on the children, on the children in their classes, critical thinking.

On the one hand, they're using it themselves and they say that it's going to increase its productivity. On the other hand, they've seen kids cheating. They're worried about it taking their critical thinking. There's this. Yeah, it's like, this is great for me, but I don't want you to mess with it. I think I would love to see being taught in schools. and I don't know if it is or it isn't,

but a productive approach to using AI instead of it being this easy button, this way to offload critical thinking, how about the way to exponentially catalyze rabbit holes of curiosity and really help you get to the things that you want to find quicker, but you're still synthesizing, you're still generating the questions. That would be great. That implies that AI has already been taught in school.

And as we learned last week, okay, in China it is, in India, it soon will be. As far as I'm aware, well, my kids don't learn about it at school. I don't know if yours actually learn about it in school. But what is it? But what is it? Is it the what, the why or the how? Like, is it what AI is?

Is it how do you use AI? Is it, what to say, what, why or how? Yeah, the why piece of it. When you're right, it's all three of those things, isn't it? Yeah, you're probably right. What did you think about this Ritech framework? This is the first I've ever heard of that, that's basically it's the responsible innovation in technology for children framework that is, is that part of UNICEF or is that,

where did that framework come out of? That's been used before to evaluate technologies. Yeah, established by UNICEF. Yep. So there's eight factors in that framework, safety, and security, autonomy, diversity, equity, and inclusion, competence, identities, creativity, relationships, and emotions. Ticks a lot of boxes. Yeah, without thinking too deeply, I don't think anything's missing from that. Can you know what's missing off that list, if anything? I'd like to spend a little time kind of talking through some of these that stood out to me as it relates to AI use, AI concerns. The one that really, really stood out to me is relationships. So quote in this chart here, when children work with traditional art materials, they generally did so while chatting

with their classmates. They're more engaged in what's around them physically if they're doing something that's outside a computer. And if they're if they're locked into doing something on a computer, they're locked into the computer or the device or whatever it is versus engaging with their peers. And that gave me a little pause. What did you think about that piece of it? The kids were divided, weren't they? There was some of them really enjoyed how it helped them with their creativity and their art. Some of them resented it. One of them said they were jealous of AI because it was better than them at art. Let me just bring up some of the quotes on the art section from the kids. That's kind of tough to hear right off the bat. Children often stated that they

enjoy the experience of physically creating artworks and using more tactile media, which is what you said. I like to feel the droplets of paint on my art brush. I like coloring and sticking things on myself. I did not like using AI. I liked making crafts without generative AI. I like doing art a bit more. I prefer doing things physically. I don't want to use a laptop. The experience feels a lot better when you're physically doing it. I felt art materials were better, slower, but it ended up what you wanted. to be. And then some very powerful quotes and insights from the kids, which you would expect to come from adults and you've probably heard adults saying a lot of these. AI can create an image that gets quite close to your imagination, but it never really gets there. No, AI is too easy. AI is like

cheating. Then another one I do not like drawing so much, so I'm quite happy about the output. The generative AI created of me. It can make a lot of stereotypes around people, so it is generating all men, all white. They all look exactly the same. It only came up with an image of a white person and it would not change. It made me feel annoyed and sad. The images need to be more diverse. Generative AI needs to be less biased. If you don't have kids, they're very smart.

They're much, much more tuned in than we often give them credit for. These kids in this survey were very concerned about the environmental impact of AI straight away as soon as they became aware of that. To the point where some of them didn't want to use it. Quote, it is ridiculous that it uses so much water. Quote, I think people. should use it less if it has the impact that it has on the environment. Quote, guilty because it is so bad for the environment.

They're aware of that. They speak about the trust layer. They say the AI made it. It's not me. Sometimes it works okay. Sometimes it doesn't do what I asked it for. This could affect you because you wouldn't know the real thing. So nods to conscious AI, seemingly conscious AI.

Definitely encouraged by the awareness of these children of the bigger picture. Sure, and I think it's largely because they're unencumbered by the bullshit that turns our thinking into path A or path B. You get hardened in your beliefs, you get less open-minded by relying on shortcuts over time. You just, you find yourself in this one group versus this other group. No matter if you share some characteristics of the other group, the easy button is to go into one group. And that stuff solidifies, it calcifies. and it's very hard to change when you get old. And they don't have that.

They're not encumbered by that. They're free from it. And like Pablo Picasso said, every kid's an artist until it gets pushed out of us. I'm loosely framing that. But man, can AI help kids stay kids? All of these are real quotes from 9 and 11 year olds in the survey. And they wrote them all down on pieces of paper.

So you can see them in the documents. On chatbots, they were saying, quote, computers don't have any feelings, why would you talk to them? No privacy in AI chat. Anyone could look at that. I think that people who make generative AI should do something to protect it somehow. They speak about misinformation, disinformation, fake news, on fake news. A nine-year-old said, misinformation could be fake news. When people start using AI, they could use it for their own purposes by making fake information and putting it online to make money. And I want to stop on that one. And before we move on to the recommendations from the Institute on what we should be doing,

did a 10-year-old really say that? Or were they just parroting what their parents have said? That sounds like me. That sounds like my kids have gone to school and said exactly what the parents are saying. And that's very important. And I think that goes back to the private school and the public school usage. I think it goes to how confident and confident children are, how confident teachers are, how worried parents are, mirrors how worried the kids are. Parenthooding is a part of learning. Here we go, Jeremy, with his musical reference again, but, you know, how do we learn how to play guitar? We learn other people's songs, and we learn to perform those songs and play those songs, but then we get our own ideas, you know, from that. So I think whether it's a parroting,

I think it's, I think it's a parroting of important messages. So good job parents of those kids to get them thinking in such a way. Open AI is always talking about needing a refresh of their board, they get some damn kids on it and get them on the right track. So recommendations from the survey. The biggest surprise for me was exactly that to get the kids involved in the development, the ideation, the production. If the kids are going to be using these tools to augment their thinking, they have to be child-friendly. They should be child versions of chat GPT.

They should be chat, GPT, 4, 5, 6 and junior, or whatever it is. And the kids should be involved in development. and designing those. And I thought that was inspired recommendation. Because a lot of the other recommendations were very obvious and wrote. And I didn't really see anything new there. The one thing I wanted to touch on was, let me see if I can find it. There was a stat in here about use of AI related to those with, I don't know how they

phrased it, not learning disabilities. Learning difficulties. Learning difficulties. They were using it a lot more than. the kids who didn't have learning difficulties, which makes sense. The thing that I thought about with that is the survey pointed to those particular individuals leaning on that more as a friend, as someone that can be a confidant for them, as they're trying to figure things out.

I got a little scared with that. I got a little sad about that because those with difficulties, learning, making friends and that sort of thing, those are the people that need more human time to help them figure it out. But I think people will lean on an easy button of like, oh, well, just let the kid use this thing to solve that harder part of life. And then they get into this loop of, oh, man, is this thing conscious? Is this thing my friend? And you get down, get down that whole rabbit hole. So I think, I think it's important to call that out and make that not be. an easy button for the path of some of these kids that actually need more attention.

I understand that certain cases, and maybe there's a supervision thing that can happen with that, but that freaked me out a little bit. I think you're right to be a little bit worried and concerned about that, again, as seems to be the case with all of the AI talk, room for good, room for bad. So recommendations, the project outlines about eight recommendations for parents, teachers, policymakers. Number one, promote child-centered AI and children's participation, which is what we just alluded to. I'm not sure what that would look like. Yeah. And it's also with these recommendations, who's going to do it? Like, who's going to take the hard job of doing it? You know, because from AI,

open AI, going to get a load of kids into their boardroom, like you said, and beat out a humane chatbot. This is all out. This is all outside in approach rather than, rather than inside out, because inside out, the race is, more GPUs get to AGI the fastest. So what is easier is develop age-appropriate generative AI tools. That would be something that the big AI firms could do, and they could do it pretty easily, much more limited data, data sets which are child content. I mean, there wouldn't be a lot of money in it, would there?

There could be. They could monetize it if it went into the schools, so that it could be a money spinner. But not a big one. So it's not a profit-driven thing. It's a humanity-driven thing. The equations, the equations need different symbols. Okay.

Support offline and online. You like this one. I like this one. So when advancing AI develop approaches with support, which support wider infrastructure simultaneously ensuring generative AI adds value alongside, not instead of more tactile materials and approaches. I love that. And what that could look like is if you're engaged on a chatbot platform after a while,

Maybe it's go outside, touch grass, get out in the sunshine. Be nice if like social media stuff would have that as well and just be like, I'm closed the curtain and you have to go outside now. My diverting are tangible takeaways. One of the kids they quoted this, and I've written here, shape AI by environment and context, not everything can be improved and made better with an AI hammer. So you call it the shortcut button. People just want to solve all their problems with AI. AI is the solution to all of these challenges and all of these creative endeavors. And it's not.

And I like this idea of building alongside that. But again, what might that look like at home or at school? Number three, improve AI literacy. Let's improve literacy in general. Would help would help that cause. How do we improve literacy in general and AI literacy? How do you push that out to all of the millions of schools in the world? Man, schools are systems that are very tough to integrate new stuff into.

If we just look at it from our particular local perspective, it's really difficult to get new information into old structures like that. Trying to figure out how to do that would be step one just in general and then figure out the message of how do you improve AI literacy into that model. It's a step one before step two. One of the big statistics, and this was 80% of teachers are using AI, but 70% of those use their personal accounts because the schools don't have. Right. The schools aren't giving them access to AI yet.

I think it's a risk thing. I think it's a risk mitigation. Number five on the list of recommendations is equitable access. And so perhaps equitable access doesn't just mean for schools. It means for business and society at large. Of all of the smaller, less financially viable corporations not having their subscription, are charities not having subscriptions, schools aren't having subscriptions, and the only people are having subscriptions are the big tech companies

or the big commerce platforms, big banks, the haves and hafnots everywhere, everywhere. It's a snowball and it's gathering speed. And a lot of these recommendations and a lot of these statistics are, I don't want to say set in stone, but I fear that if we came back, when we come back in a few years, that the numbers will have grown, but the disparances will remain. The gaps are the gaps. The gaps are the gaps and the gaps will remain the gaps.

I really hope policymakers and maybe some people in UK government in the education department of the UK government have read this and are acting on this and are doing something with this with the boards of teacher boards and hopefully acting on it and making AI more equitable, fairer, more distributed and a nice experience for the kids who are going to be using it in school. What other papers guys? Because they all done to be using it.

Yes. What other papers do we need to read in research and on past? What do you not want to read that you want to hear us talk about? We'd love you to read it because that's how you make it happen. And you can even write about it. It's called Thinking on Paper. But if you want to listen to us, do it. We're going to do it anyway because we like to understand this stuff.

So put something in the comments. Let us know what to tackle next. Empire of AI. We're still unpacking it. We got one more part to do in that series. Stay tuned there. Got some great guests coming up. A closing thoughts, Mark, before we get out of here.

We are tech agnostic. We speak about AI. We talk about space. We talk about quantum computers. computing robotics, human or bioengineering, doesn't matter what it is. If you've got a paper, if you've got an idea, if you've got a conversation, put it in the chat. And we'll spend some time thinking on paper about it. Thank you for joining us.

Check out thinking on paper. comyZ. And we'll see you next time. Be curious. Stay disruptive. Keep thinking on paper.