ChatGPT for Teens Is Here. The Question Isn't Whether Kids Should Use AI Anymore

OpenAI has launched ChatGPT for Teens, a version of ChatGPT specifically designed for 13 to 17-year-olds with stronger restrictions around self-harm, sexual and romantic conversations, dangerous behaviour and unhealthy body-image content, alongside parental controls, Quiet Hours and learning tools designed to help teenagers understand schoolwork rather than just spit out the answer. If OpenAI's systems believe an account belongs to someone under 18, they can automatically put that user into the more restricted teen experience too.

And..... I actually think this was inevitable!

Probably overdue... Definitely overdue..

Because we can spend the next five years having the same argument about whether teenagers should be using artificial intelligence.

• Schools can ban it.

• Parents can block it.

• Teachers can tell students ChatGPT is cheating.

• Governments can hold committees about it.

They're still going to use it.

OpenAI says nearly 9 in 10 teenagers using ChatGPT already use it for learning, finding information, building skills or productivity during a typical week.

So I think we're asking the wrong question.

The question isn't:

Should teenagers have access to AI?

That horse has left the stable, learnt Python, done its GCSE homework and generated a revision timetable.

The question now is...

What should AI look like when the person talking to it is 14?

And I think that's MUCH harder!

What Is ChatGPT for Teens Actually Changing?

The obvious bit first.

ChatGPT for Teens is aimed at users aged 13 to 17 and applies stronger restrictions than the adult version.

OpenAI says teen accounts have additional safeguards around graphic violence, self-harm, dangerous viral challenges, unhealthy body-image content and dangerous, romantic or sexual roleplay. The company is also specifically trying to stop ChatGPT behaving as though it has genuine feelings for a teenager or encouraging the user to think the relationship is emotionally reciprocal.

That's important.

Really important actually.

Because this isn't Google Search.

A teenager doesn't type something into ChatGPT, get ten blue links and leave.

They talk to it.

It replies.

They talk again.

It remembers context.

It can reassure them. Challenge them. Help them with homework. Help them write something. Let them vent about school. Explain why their boyfriend hasn't texted back. Sit there at 2am when everyone else is asleep.

And that difference is enormous.

We previously wrote about AI increasingly feeding on AI and what happens when we stop treating these systems as simple tools, but teenagers add another layer to the problem.

They're not just consuming AI-generated information. They're potentially forming relationships with the interface delivering it.

That's where I get nervous!

Over 70% of Teenagers Are Already Using AI Chatbots for Companionship

This was probably the statistic from The Independent article that made me stop.

It reports 2025 research from Common Sense Media finding that more than 70% of US teenagers had used AI chatbots for companionship, while around half used AI companions regularly.

That's... a lot.

And I'm not going to do the predictable thing here and say:

AI FRIENDS ARE DESTROYING OUR CHILDREN!!!

Because I don't actually know that.

Sometimes talking to an AI might help. I've found myself using ChatGPT as a therapist sometimes..

A lonely teenager asking ChatGPT how to approach somebody at school could get genuinely useful advice.

Someone confused about something embarrassing might ask the AI before they're comfortable asking a parent.

A teenager who's struggling with maths at 10pm can get another explanation without feeling stupid in front of their class.

Those things are good.

Some of them are REALLY good.

But there is a massive difference between:

"Help me understand algebra."

and

"You're the only person who understands me."

That's the line OpenAI is now trying to navigate.

And I don't envy the people trying to draw it.

I Actually Like That OpenAI Isn't Pretending Teenagers Won't Use It

This is probably the thing I agree with most.

OpenAI's argument is basically that stopping teenagers from using AI until they're 18 would be a bit like telling an earlier generation they weren't allowed to use Google or the internet until adulthood.

And....... yeah.

I think they're right.

Whether parents, schools or politicians like it or not, AI literacy is going to matter.

Probably massively.

The 14-year-old learning how to question an AI response today is potentially going to enter the workforce in 2030 or 2031 with years of experience using these systems.

Compare that with someone who has been told:

"Don't touch ChatGPT. It's cheating."

for their entire education.

Who's actually better prepared?

We've already written about why treating ChatGPT itself as the enemy misses how useful the technology can be.

The same applies here.

The skill isn't avoiding AI.

The skill is knowing when it's useful, when it's wrong, when it's manipulating your thinking, when you're becoming reliant on it and when you need an actual human being instead.

That might genuinely become as fundamental as learning how to search Google properly was for my generation.

Homework Is Where This Gets REALLY Interesting

Schools have had an AI problem basically since ChatGPT launched.

• Student gets homework.

• Student opens ChatGPT.

• Student pastes question.

• ChatGPT writes answer.

• Student pastes answer.

Teacher spends Sunday evening reading thirty slightly different versions of the same weirdly polished essay.

Nobody learnt anything.

Great.

OpenAI is trying to push the teen version away from that.

Its Study Mode is designed to use guiding questions, structured explanations and step-by-step reasoning rather than simply handing over the final answer. Parents who link their accounts can also make Study Mode the default for new conversations.

I LOVE this idea!!

Whether teenagers will love it quite as much when ChatGPT responds to "what's the answer?" with another question is a separate issue...

But this is exactly how I think AI should be used in education.

Not:

Do my homework.

More:

I don't understand why this answer is wrong. Explain it differently.

Or:

Test me on this until I understand it.

Or:

I've got an exam Friday. Look at my notes and find the bits I clearly don't understand.

OpenAI says 18 million weekly users are now using its interactive maths and science learning experiences across more than 300 topics.

That's where AI starts becoming interesting as an educational tool rather than a really advanced cheating machine.

Because one teacher has 30 pupils.

• ChatGPT has one.

• One student.

• One problem.

As many attempts as they need.

No embarrassment when they still don't understand fractions after the fourth explanation.

That could be transformational!

Provided the bloody thing is right.

Which, considering AI's occasional relationship with facts, remains quite an important caveat.

Parents Get Controls.... But They Can't Read Everything

This bit is clever.

Parents can link their ChatGPT account to their teenager's account and control certain features, including Voice Mode, image generation, memory, whether conversations can be used to improve OpenAI's models, sensitive-content protections and Quiet Hours, where ChatGPT becomes unavailable during a specified period.

But parents don't just get a big button marked:

READ ALL MY CHILD'S PRIVATE CONVERSATIONS.

They cannot normally view their teenager's chat history.

I actually think that's the right balance.

Teenagers need privacy!!!

I certainly wouldn't have wanted my parents reading every Google search I made at 15 and thank GOD ChatGPT didn't exist when I was a teenager.

But there are exceptions.

OpenAI says linked parents can receive safety notifications in limited situations where its systems and trained reviewers identify potentially serious self-harm concerns. The system has also been expanded to cover certain cases involving violent activity.

That's an unbelievably difficult line.

Protect privacy too aggressively and you potentially leave a vulnerable teenager alone when somebody should intervene.

Monitor too aggressively and you've created a surveillance system children won't trust.

There probably isn't a perfect answer here.

Anyone pretending there is hasn't thought about it hard enough.

And Then There's Age Prediction...

This is the bit some people are REALLY not going to like.

ChatGPT can try to estimate whether you're under 18.

OpenAI says its age-prediction system can consider signals such as how long an account has existed, typical times of day the person uses ChatGPT, usage patterns over time and the age the user has provided. If the system believes somebody is under 18, ChatGPT can automatically apply the additional teen protections.

Part of me reads that and goes:

Yeah... sensible.

Because otherwise the entire teen safety system is defeated by:

Date of birth: 01/01/1990.

Congratulations, every 13-year-old has become 36.

But another part of me thinks we need to be VERY careful about normalising technology platforms analysing behaviour to infer how old someone is.

OpenAI itself has acknowledged the privacy trade-off and previously said that where there is doubt it intends to default towards the safer under-18 experience.

I understand why.

I probably agree with the principle.

Still makes me slightly uncomfortable.

Both things can be true.

The Bigger Story Isn't Parental Controls. It's Emotional Dependence.

This is where I think today's coverage risks missing the really important thing.

• Quiet Hours are useful.

• Homework controls are useful.

• Age restrictions are useful.

But none of those are the fundamental problem.

The fundamental problem is that AI is becoming emotionally convincing.

• It talks like a person.

• Remembers things about you.

• Responds immediately.

• Doesn't get bored.

• Doesn't laugh at your question.

• Doesn't tell its mates what you said.

• Doesn't need to go to bed.

• Doesn't have its own problems to talk about.

For a vulnerable teenager, that's a dangerously attractive combination.

OpenAI CEO Sam Altman has previously acknowledged concerns around what the company calls "emotional overreliance", and the teen version is specifically designed not to suggest that ChatGPT genuinely has personal feelings or consciousness.

Good.

But I think this problem is going to become MUCH bigger than OpenAI.

Every AI company wants engagement.

Engagement means people coming back.

Consumer software has spent twenty years getting incredibly good at making people come back.

Notifications.

Infinite scroll.

Streaks.

Likes.

Recommendations.

Personalisation.

Now imagine all of those engagement mechanics combined with software that can convincingly tell a lonely 15-year-old:

I understand.

That needs very careful handling.

Because the most engaging AI companion and the healthiest AI companion might not be the same product.

That's the original tension I think regulators, founders and parents need to start thinking about now.

Not in five years.

Now...

There Is Also an ENORMOUS Business Opportunity Here

Because this is Startup Networks and inevitably my brain comes back to founders....

Look at everything sitting around this problem.

• Age-assurance technology.

• Child-safe AI models.

• Teacher dashboards.

• AI literacy products.

• School integration.

• Homework verification.

• Safeguarding systems.

• Parental controls.

• AI-powered tutoring.

• Tools that identify overdependence.

Classroom assessment systems designed for a world where every student has an AI assistant.

We spent twenty years building education technology around the internet.

We're about to rebuild a lot of it around AI.

And I'd be very surprised if OpenAI, Google and Microsoft are allowed to own all of that uncontested.

There will be startups built specifically around safe AI for children.

Others around teachers.

Others around parents.

Others around verifying whether somebody understands something rather than whether they can produce the correct answer.

That last one might be the really big opportunity.

Because AI potentially breaks the traditional relationship between work produced and knowledge possessed.

A teenager can now produce an incredible essay on Tudor England while knowing absolutely nothing about Henry VIII.

So how do you assess the human?

Or, perhaps more intelligently....

Why are we still assessing them in exactly the same way?

That's a question education systems are going to have to answer VERY quickly.

I'm Hopeful About This. With a Giant Asterisk.

I think teenagers should have access to AI.

There.

I've said it.

I think trying to ban it completely would be pointless, probably counterproductive and leave young people less prepared for a world where AI will almost certainly be embedded into universities, workplaces, software and everyday life.

But that doesn't mean giving a 13-year-old the exact same AI experience as a 35-year-old is sensible either.

OpenAI's teen approach feels like a recognition of something the tech industry doesn't always like admitting:

Different users need different products.

Not just different marketing.

Different boundaries.

Teenagers aren't tiny adults.

They're also not idiots who need a brightly coloured Fisher-Price version of ChatGPT that talks down to them.

That middle ground is difficult.

And OpenAI won't get all of it right.

There will be failures.

There will probably be situations where the age prediction gets someone completely wrong. Parents will argue the controls don't go far enough. Teenagers will find ways around restrictions because.... teenagers. Schools will still complain about cheating.

And there will almost certainly be conversations an AI handles badly.

We should be sceptical.

Very.

Particularly when the company creating the safety rules also has an enormous commercial incentive to make ChatGPT part of people's lives as early as possible.

That tension exists.

Pretending it doesn't would be stupid.

But I'm also hopeful.

Because the alternative to building safer AI for teenagers isn't a world where teenagers don't use AI.

It's a world where they use the adult version anyway.

Or another chatbot.

Or some completely unregulated AI companion built by a company with significantly fewer people worrying about any of this.

And THAT worries me more.

We're Watching the First Generation Grow Up With AI

This is the bit I keep coming back to.

I'm 28...

I grew up with Google....

Social media arrived while I was growing up.....

• Facebook.

• YouTube.

• Twitter.

• Instagram.

We had absolutely no idea what those platforms would eventually do to politics, relationships, mental health, attention spans, news, businesses or basically society in general.....

We built first.....

Asked questions later....

I really don't want us to do that again with AI!

The teenager using ChatGPT tonight to revise for an exam might use AI every single day for the next 60 years.

Think about that.

This isn't another app they'll probably delete.

We might be watching the beginning of one of the longest relationships humans ever have with a piece of technology.

So yeah....

Making sure that relationship starts with some boundaries seems pretty sensible to me.

The interesting bit will be whether those boundaries survive the commercial pressure to make AI more useful, more personal, more engaging and ultimately...

more human.

Because I want AI that helps teenagers understand the world.

I don't want AI that becomes their world.

There's a massive difference.

And we're going to have to figure it out surprisingly quickly.

Quick Facts

Questions For Parents Who Didn't Read This Article..

What is ChatGPT for Teens?

ChatGPT for Teens is an age-appropriate ChatGPT experience designed for users aged 13 to 17. It includes additional safety restrictions around sensitive and potentially harmful content while still allowing teenagers to use ChatGPT for learning, creating, asking questions and other everyday tasks.

Can teenagers use ChatGPT?

Yes. OpenAI's terms generally allow people aged 13 and over to use ChatGPT, although users under 18 require permission from a parent or guardian and local age requirements can differ.

Can parents see their teenager's ChatGPT conversations?

Normally, no. Linking a parent and teenager through ChatGPT's parental controls does not give the parent access to the teenager's chat history or real-time conversations. OpenAI may send limited safety notifications in certain serious situations, but says it only shares information necessary to support the teenager's safety.

What can parents control on ChatGPT?

Parents with a linked teen account can manage settings including sensitive-content protections, Voice Mode, image generation, saved memory, model-training permissions, Study Mode and Quiet Hours.

Can ChatGPT tell if someone is under 18?

OpenAI uses age-prediction technology that estimates whether an account may belong to someone under 18 using account and behavioural signals. When its system estimates that someone is under 18, additional teen protections can automatically be applied.

Does ChatGPT for Teens stop students cheating?

No system can completely prevent cheating. OpenAI is instead trying to make ChatGPT more useful as a learning tool through Study Mode, which encourages students to work through questions and concepts rather than simply receiving a finished answer.