The honest answer is that nobody knows yet, and the people best placed to say so admit it. But the comparison is still worth making, because these are not the same problem. Social media mostly reached your child through other people. AI reaches them through what it offers to do for them, and who it offers to be. That difference is what should change your approach at home.
One thing before the rest. If your worry right now is not AI in general but your own child, something they have said, or written, or stopped doing, then that is a conversation with your doctor or your school's safeguarding lead, and it should not wait until you have finished reading about technology.
What the people sounding the alarm actually said
Tom Siegel ran trust and safety at Google for fifteen years. On September 22, 2026 he became executive director of a new Youth AI Safety Institute at Common Sense Media, and told Reuters that "we find ourselves in a really terrible situation in terms of what AI is doing to kids," and that the idea the industry "will figure it out by itself without outside help is unrealistic." That was syndicated everywhere within about two days, which is probably where you met the headline.
He is not the first to say it. At a US Senate Commerce Committee hearing on January 15, 2026, Jean Twenge, a psychology professor at San Diego State University who has spent years studying teenagers and phones, told the committee: "I'm even more concerned about the AI companion apps than I am about social media." Jenny Radesky, a pediatrician at the University of Michigan, gave the reason: "My biggest concern is attachment and relationships. Kids are wired to want to attach to other humans...the AI companions are exploiting this."
Now notice what those statements are, and what they are not. They are expert judgments from people with real track records, and they are worth taking seriously. They are not findings. Social media has around fifteen years of research behind it and researchers still argue about what it shows. Consumer AI has had about two. Anyone who tells you with confidence that AI is worse, or that it is fine, has gone further than the evidence has.
The difference that matters more than the ranking
So if the ranking cannot be settled, what is the comparison actually for? This, I think.
Social media's harm ran through other people. The pile-on, the group chat your child was left out of, the photograph, the number of likes under somebody else's name. The machinery was social, which is why the advice that came out of it was social too: delay the account, know who they are talking to, keep the phone out of the bedroom overnight.
AI's route runs the other way. It works by substitution. It offers to do the piece of thinking your child was about to do, and it offers to be the company they were about to go looking for. Both feel like help in the moment. That is exactly what makes them hard to spot from the outside.
The measurements we do have point at both halves of that. Common Sense Media's census of AI use by tweens and teens, published on June 8, 2026 from a survey of 1,204 children aged 9 to 17 in the United States, found that more than one in three of the children who use AI have used it to talk about their feelings or personal problems, and that one in five say it would be hard to go a month without it. In the UK, the Internet Matters study "Me, Myself and AI", published on July 12, 2025, found the pattern sharpest among the children it classed as vulnerable: half of them said talking to an AI chatbot is like talking to a friend, and 26 percent said they would rather talk to a chatbot than to a real person.
Read those carefully, because they describe an association and not a cause. A child who talks to a chatbot because they have nobody else quite possibly had nobody else first. That does not make the finding useless. It makes it a signal about which children this reaches hardest, which is the more useful thing for a parent to know.
Why the social media playbook only half works here
If you parented through the social media years you built a playbook: delay the account, restrict the device, no phones at night, and talk about what happens in the group chat. Schools built the same one, and most of them eventually landed on the pouch or the locker.
Half of it transfers. The talking half transfers completely. The restricting half does not, for two reasons.
The first is that AI is not a destination. Social media was a place your child went, and a place has a door you can stand in. AI is a feature inside things they already hold: the answer above the search results, the assistant on the phone, the chat pinned at the top of Snapchat, the "rewrite this" button in the app they type their homework into. There is no account to delay when nobody had to open one.
The second is the one parents find hardest. The school is handing it to them deliberately. Whatever you settle at your own kitchen table, your child will meet AI in a classroom because an adult in that building decided they should, and in most schools nobody has written to you to say what that decision was.
With social media, the question was whether to let it in. With AI, it is already in, and the question is what your child hands over to it.
What I See in Schools
I speak about this in a lot of countries, and the shape of the worry is the same in all of them. The question underneath is never really "is it safe." It is "am I getting this wrong, and will I find out too late."
I was working in schools through the social media years, and the lesson I took from that period is not the one usually drawn. Schools did act in the end, and phone bans do help with the thing they are pointed at. But the years before that were spent arguing about whether the problem was real, and the children who were in school during those years did not get them back.
The pattern I see most often now is a school that has made a decision about AI and has not told parents that it made one. Rarely out of secrecy. Usually because the decision was taken by two or three people in a room with a hundred other items on the agenda, and writing to parents never got onto anybody's list.
What to do about it this week
Stop measuring time and start asking what the time replaced.
Screen time was a fair proxy for social media, because time on the app was close to the harm itself. It is a poor proxy here. Twenty minutes in which a child argued with a chatbot about a history essay and came out with a better argument is not the same as twenty minutes in which it wrote the essay, and no clock can tell the two apart.
So tonight's useful move is a question rather than a setting. Pick one specific thing they used it for and ask what they did with what came back.
You could say: "What did it get wrong?"
That is close to what the American Academy of Child and Adolescent Psychiatry recommends in its guidance for families on AI, published in July 2025, which asks parents to "explore AI tools together with your child" and to "model curiosity and skepticism: teach children to question information from AI." That question does more than it looks like it does. A child who has been thinking alongside a chatbot will have an answer, usually a slightly indignant one. A child who has been handing things over will not have noticed anything to be wrong.
And if the part that worries you is the company rather than the schoolwork, you could say: "Is there anything you'd say to it that you wouldn't say to me? I won't be upset. I'd just rather know."
The principle underneath both is the one I give to schools, and it holds at home: let AI help with the doing; keep the thinking with your child.
How this changes by age
Under 8. The comparison hardly applies yet, because your child is not on social media and should not be alone with a chatbot either. What they will meet is the assistant that answers instantly, on a speaker or a tablet, and the habit worth protecting is the pause before the answer. You could say: "What do you think it's going to say?"
8 to 12. Substitution usually starts here, with homework, while the social media playbook is still worth running because the accounts are still ahead of them. Do both. The Internet Matters study found 58 percent of 9 to 12 year olds in the UK already using AI chatbots despite minimum ages of 13, so "not yet" and "not happening" are not the same sentence.
13 to 15. Both problems are live at once, and this is the band where emotional disclosure starts showing up in the data. The most useful thing you have here is not a control. It is a standing agreement that they can tell you when something a chatbot said unsettled them without it costing them the device.
16 to 18. Controls are mostly the wrong tool by now. What matters is whether they can tell a machine that is helping them think from one that is thinking instead of them, because they are about to carry that into exams, applications and a first job.
What none of this can fix
Three limits worth being straight about.
No setting separates help from substitution. Parental controls can block an app, hold an age rating or switch off a named assistant. Not one of them can tell you whether your child understood the paragraph. That judgment sits with a person, which in practice means you and your child's teacher.
Your child's school sets the rules for its own devices and its own work. Where its rule and your rule disagree, the school's applies at school, and the way to settle it is a conversation with the school rather than a setting at home.
And the alarm in this week's headlines is about severe harm, including children in genuine mental health crises. If that is where your worry actually sits, this is not a technology decision. In the United States, the 988 Suicide and Crisis Lifeline takes calls and texts. In the UK, Samaritans are on 116 123 and Childline is on 0800 1111 for under-19s. Your family doctor and your school's designated safeguarding lead are the people who can act, and a chatbot is not a clinician however it sounds.
What I would do this week
Pick one thing your child used AI for in the last seven days and ask them what it got wrong. That is the whole exercise, and it tells you more than a week of screen time reports.
Then ask the school. If your child's school has not written to parents about what it has decided on AI, that is the gap worth closing, because school is the one place you cannot opt out on your child's behalf. Working out what a school has actually decided, and saying it out loud to parents, is the readiness work I do with schools, and it is worth sending them.
I write a weekly email for parents and teachers on this, without the panic. You can sign up here.
Dan Fitzpatrick is a former teacher and assistant headteacher who now works with schools on AI. More about Dan.
