Is AI bad for kids? What the research actually shows

Short answer: yes and no, and the difference between the two is entirely in your hands. Unguided AI is bad for children. It does their thinking, it tells them confident nonsense, and it fills the hours that used to hold boredom and people. Guided AI is a different thing altogether: a patient tutor, a sparring partner for the mind, and a way to make things a child could not make alone. The research does not say AI is good or bad. It says the way it is used decides.

I have read most of that research, because I had to. I have children who reached for these tools before I had an opinion about them. Here is what it actually shows, worry by worry.

The gap between fear and evidence

A 2024 Barna Group survey found that 72 percent of parents are concerned about AI's effect on children and teenagers. In the same survey, only 17 percent strongly agreed that they seek out information about it, and most said they think about it rarely or never. The anxiety is nearly universal. The engagement with the details is rare.

That gap matters, because vague anxiety produces vague responses. The parent who bans every AI tool because they are worried about cheating has not prevented cheating. They have made sure it happens somewhere they cannot see. The most useful sentence I can give you is this: the problem in almost every case is not AI use, it is unguided AI use.

Worry one: cheating

Let me not pretend. A survey of over a thousand college students by Study.com found that 89 percent had used ChatGPT for homework. College Board research published in 2025 found that 84 percent of high school students now use generative AI for schoolwork. Your child is using it, or will be, whatever the school's policy says.

So the question is not whether. It is how. A child who types "write my essay on the causes of the French Revolution" and hands in the result has cheated, and worse, has learned nothing. A child who writes a rough draft, then asks the AI to find the weakest argument in it, then fixes it themselves, has done something a good teacher would applaud. Same tool. Two paths. I wrote a whole chapter about the two children in those two houses, and it is free to read.

Worry two: it stops them thinking

This is the worry with the best evidence behind it, so take it seriously.

Now hold that next to a different study. A Harvard randomised controlled trial found that a carefully designed AI tutor more than doubled students' learning gains compared with an actively taught class, and the students were more engaged and more motivated. Same technology. Opposite result.

The difference was the mode. In the Harvard trial the AI was built to make students think: it asked, it hinted, it checked understanding. In the MIT study the chatbot was simply asked to produce the essay. When AI is a thinking partner, cognition improves. When AI is a replacement thinker, it atrophies. That is the whole finding, and it is the reason this website exists.

Worry three: privacy

Everything a child types into an AI tool goes to a company. Some companies are careful with children's data and some are not, and a child cannot tell the difference. The fix here is a rule, not a fear: only put into an AI what you would happily show a stranger in the street. Names, addresses, photos of friends, the thing they are ashamed of: none of it. And for younger children, use tools built for minors where they exist, because a children's version with real content filtering and data limits is a fundamentally different product from a general assistant.

Worry four: screens and people

A majority of parents worry that AI will replace human connection in their children's lives. The screen research is genuinely mixed and most of it predates today's AI, but one pattern is well established: displacing human connection with device time, especially passive, solitary consumption, is associated with poorer social development. Adding device time on top of a full human life shows much smaller effects.

So the question for your family is a question of balance, and you can answer it tonight. A child who uses AI tools for an hour after dinner and then spends the evening in conversation, play and family noise is not at meaningful risk. A child who is alone in their room with AI most evenings is. Human connection is not protected by restricting AI. It is built by investing in it: dinner without phones, protected time to be bored, a question jar on the table.

Worry five: the jobs

Parents ask me this one quietly, at the end, when the others have left. Will there be anything left for my child to do? I have spent years watching AI reshape real companies, and my honest answer is that the work that survives is the work that is most human: judgement, care, imagination, the ability to sit with a hard problem and not give up. Those are skills, and they can be built at home from the age of five. The five I built the book around are resilience, curiosity, adaptability, systems thinking and empathy. Not because AI can never touch them, but because they are the ones worth prioritising in a world shaped by it.

One more number

In 2024, 64 percent of parents supported using AI to help children prepare for tests, and 65 percent supported AI tutoring. A year later those figures had fallen to 54 and 60 percent. Trust is going down. That decline tracks the volume of alarming headlines, not any measured increase in harm. Parents are getting a steady diet of fear without an equally accessible diet of evidence. Consider this post a small correction.

What guided looks like at home

You do not need a computer science degree. You need three rules and one habit.

The 20-Minute Rule: before asking an AI anything about schoolwork, your child works on it alone for twenty minutes. Struggle first, then help.

The V.E.T. rule: anything an AI produces gets Verified against a second source, Explained back in your child's own words, and Transformed into their own work.

The Family AI Agreement: one page, written together, saying which tools, for what, what never gets typed in, and what happens when something goes wrong. The builder on this site is free.

And the habit: protected human time that the tools cannot touch. Meals, walks, boredom, bedtime.

Questions parents ask

Is AI bad for a child's brain?

The best current evidence, a 2025 MIT study, found that using a chatbot to produce whole essays weakened engagement, memory and ownership of the work. The same body of evidence finds that AI used as a tutor or thinking partner improves learning. The tool does not decide; the mode of use does. Keep your child doing the thinking and the risk largely disappears.

At what age is AI safe for kids?

Together with a parent, for play and stories, from about five. Supervised on a shared device, with the 20-Minute Rule, from about nine. Independently, with the V.E.T. rule and a written family agreement, from the early teens. Companion-style chatbots are a different case and I would hold those until the teens at the earliest.

Should I ban AI at home?

For most families, no. A ban removes you from the conversation without removing the tool from your child's life. The parents who do best are the ones who set rules about how AI is used and stay curious about what their child is doing with it.

Does AI make kids lazy?

It can, if it is allowed to do the work. It does not have to. A child who asks an AI to challenge their draft, explain a concept a second way, or quiz them before a test is working harder, not less. The habit you want is "help me think", never "think for me".

Sources

• Barna Group, 2024, survey of parents on AI and children.
• Study.com survey of college students on ChatGPT use for homework, 2023.
• College Board research on high school students' use of generative AI, 2025.
• MIT Media Lab, 2025, the essay-writing and brain-activity study reported as "Your Brain on ChatGPT".
• Harvard University randomised controlled trial on AI tutoring.
• Parent-trust figures for 2024 and 2025 as cited in Chapter 17 of Raising AI-Ready Children.