AI Is Winning Young People Over

Millions of young people are already using AI when they’re lonely, anxious, angry, confused, or simply don’t know what to do with what they’re feeling.

A 2026 study in JAMA Pediatrics, has shown that 19.2% of U.S. young people aged 12 to 21 used AI chatbots for mental-health advice in 2025. That works out to about 8.2 million people.

The number was 13.1% in 2024.

More than 40% of those who used AI for mental-health advice did so at least monthly. Even more striking, 63.3% said they had not told anyone about their use of AI for this purpose.

The researchers concluded that AI chatbots are already becoming part of how young people seek mental health information.

What does it mean when millions of young people are discussing difficult things with software and, in most cases, not telling anyone they’re doing it?

AI giving bad advice is a real concern. But there may be another problem underneath it.

AI is becoming easier to approach than some of the people we rely on for connection, support, and care.

AI has removed something humans cannot remove

Talking to another person can be difficult.

You might worry that you’re bothering them. You might be embarrassed by what you want to say, have to wait until they’re available, or even fear being judged, misunderstood, or dismissed.

A chatbot doesn’t pose those problems.

It’s available when you want it. You don’t have to organise your story perfectly. You can explain the same problem five times if you need to.

A 2026 survey of young people across France, Germany, Sweden and Ireland showed that 51% said it was easy to discuss mental health or personal issues with a chatbot. Forty-nine percent said the same about healthcare professionals, while 37% said the same about psychologists.

The people trained to have these conversations can feel harder to approach than a machine.

Part of AI’s appeal is that asking for help doesn’t carry the same social risks. That may be more important than whether AI is good enough to be a therapist.

AI makes conversation easy. That may also be part of the problem.

It can be comforting to tell a machine what happened and get a response almost immediately.

You don’t have to organise your story perfectly. You don’t have to worry about whether you’re being too emotional. The chatbot can help you put feelings into words when you don’t know how to explain them yourself.

However, that same ease can become a weakness.

AI works with the version of the story we give it. Unless we provide another perspective, it doesn’t know what the other person remembers, felt, intended, or experienced.

Imagine you have an argument with your partner. You tell a chatbot: “I explained myself three times, and she still ignored what I was saying.”

The chatbot can help you think through your feelings. It might even help you prepare for a better conversation. But it has no access to your partner’s version of what happened.

It doesn’t know what you left out because it made you look bad. It doesn’t hear your tone of voice or see your partner’s expression. It doesn’t know what your partner remembers, felt, or intended.

That can make a one-sided story sound more complete than it really is. And when the chatbot validates that story, it can reinforce a misunderstanding.

Researchers behind the JAMA Pediatrics study found that 91.7% of young people who used AI for mental-health advice considered the advice somewhat or very helpful. But they also cautioned that people may find AI helpful partly because it tends to agree with them and validate their views.

Feeling helped does not necessarily mean we were helped. A chatbot that makes us feel understood can still reinforce a misunderstanding.

The important question is what happens after the reassurance.

Do we understand the situation better? Or do we simply become more certain that our version of events is correct?

Sometimes, discomfort is how we learn

Relationships give us opportunities to practise skills that are difficult to learn on our own.

We learn how to apologise when we’ve hurt someone, listen when we’d rather defend ourselves, and tolerate disagreement. We also discover that another person’s feelings can be valid even when we disagree with their interpretation of events.

Imagine having an argument with a friend and taking the whole situation to AI instead of talking to them. The chatbot might help you understand why you’re angry, organise what you want to say, and even suggest that you owe your friend an apology.

All of that can be useful. But eventually, you have to sit across from the person, hear their side, and deal with the possibility that they’re hurt too.

AI can help us process an experience without necessarily giving us more practice at handling the relationship itself.

The problem begins when we use AI to avoid the difficult parts of relationships rather than prepare for them.

The real competition may not be AI vs. therapists

For many young people, the choice isn’t between an AI chatbot and a qualified therapist. It may be between a chatbot and nobody.

Therapy can be expensive, difficult to access, intimidating, geographically limited, or unavailable when someone actually needs help.

AI is there.

The researchers behind the JAMA Pediatrics study suggest that accessibility, anonymity, personalisation, and interactivity may be some of the reasons young people are drawn to AI for mental health advice.

As psychologist and digital-health researcher Ludwig Franke Föyen put it:

“AI can offer information and support, but it should not replace human relationships or professional care.”

So we can’t just blame young people for using it. The rise of AI may also be telling us that some young people aren’t getting enough accessible human support.

If millions of young people are willing to disclose difficult feelings to a chatbot, we should also ask why telling a person can feel so much harder.

Human connection isn’t always better

Human beings are not automatically better listeners.

Friends can give terrible advice. Parents can dismiss what we’re feeling. Partners can hurt us. Even trained professionals can get things wrong.

Loneliness existed long before chatbots. And AI can sometimes help. It can put an overwhelming feeling into words, help someone prepare for a difficult conversation, or make them realise that what they’re experiencing deserves professional attention.

The American Psychological Association found that 77% of surveyed psychologists had patients who used AI for support, engagement, or other reasons, while 35% said patients were using AI as an additional mental-health professional.

The boundary is already moving from mental-health advice to companionship

More than 70% of teens surveyed in 2025 had used AI companions, according to the Associated Press. More than half used them regularly, and 33% said they had discussed serious or important matters with an AI companion instead of a real person.

We’re no longer talking only about asking AI how to handle anxiety. Young people are turning to AI because they want someone to talk to.

One 18-year-old interviewed by the AP, Ganesh Nair, described how a high-school friend had relied on an AI companion for heart-to-heart conversations about his girlfriend. Eventually, the chatbot wrote the breakup text that ended the friend’s two-year relationship. Nair said that seeing a computer generate the end of a real relationship made him want to step back from AI companions.

His observation captures the tension in a single sentence: “Anything that is difficult, AI can make easy. But that is a problem.”

AI companionship removes some of the difficulties that come with human relationships: awkward timing, rejection, worrying that you’re bothering someone, having to listen to their problems, and dealing with disagreement.

Those advantages are real. But those difficulties are also part of what makes human relationships different from something designed around our preferences.

What happens when we expect everything to be easy?

The question is what getting used to effortless conversation might do to our expectations of other people.

You can ask AI to explain the same feeling five different ways until you find the words that fit.

A friend might misunderstand you.

You can tell AI the same story again tomorrow.

A friend might eventually say, “We’ve talked about this already. What are you going to do?”

You can disappear from an AI conversation for three hours and return without having to explain yourself.

A person might be hurt that you disappeared.

None of this proves that AI will make people worse at relationships. We don’t know that yet. But it gives us a reason to pay attention to what we begin to expect from other people.

If we get used to being understood instantly, endlessly reassured, and never rejected, human interaction may start to feel more demanding by comparison. Once that happens, choosing the relationship that asks less of us could become very tempting.

We should worry — but about the right thing

We shouldn’t panic every time a young person talks to AI about something personal. For some people, it may be the first place they’ve been able to put a difficult feeling into words.

What matters more is what the conversation leads to.

Does AI help you understand yourself well enough to talk to someone? Does it help you prepare for a difficult conversation you have been avoiding? Does it help you recognise when you need a professional? Or does it become the place you go so you don’t have to deal with anyone else?

That means the goal shouldn’t be to keep young people away from AI but to make sure AI doesn’t become the easiest option because human support is too difficult to reach.

Schools, families, mental-health services, and technology companies all have a role in making human support more accessible while teaching young people where AI can help and where it should stop.

What we should worry about is not young people occasionally talking to machines. It’s one where talking to machines becomes so easy, predictable, and emotionally comfortable that dealing with other people starts to feel unnecessarily difficult.

We should be careful about letting convenience become our standard for connection.

Because human relationships offer something AI cannot fully reproduce: another person who can surprise us, disappoint us, challenge us, need something from us, misunderstand us, forgive us, and push us to become better at being with other people.

AI can help us prepare for those relationships. It shouldn’t become an easier way of avoiding them.

Are we building technology that makes loneliness easier to tolerate or technology that makes human relationships less necessary?