Our teens are using AI chatbots daily. It's our fault. | Opinion

A teenager sits on the edge of her bed, phone screen glowing, thumbs moving fast. But she isn't texting a friend. She isn't calling a parent. She's asking an artificial intelligence chatbot what to do with the panic in her chest, the loneliness at lunch, the dread that hits at night.

That scenario isn't rare. A 2025 Pew survey found that 64% of U.S. teens use AI chatbots, and that about 3 in 10 use them daily. A 2026 JAMA Pediatrics survey found that about 1 in 5 U.S. adolescents and young adults had used AI chatbots for mental health advice. Common Sense Media also reported in 2025 that nearly 3 in 4 teens have used "AI companions," with half using them regularly.

We're panicking about AI "replacing" friendship. But the real emergency is that too many youth don't have a trusted human to go to in the first place. If a chatbot feels safer than a parent, teacher, coach, counselor or friend, it's not a technology story; it's a culture story.

We write as psychologists who have spent decades studying emotions, relationships, and youth well-being. Our work with teens, schools, families, and technology companies has convinced us that AI is not creating this crisis so much as revealing it.

Children are turning to AI because it doesn't judge them

Of course these tools need guardrails: tech company accountability, school policies, protections for minors, and serious scrutiny of products designed to simulate intimacy with young people. But if we stop there, we miss the deeper story.

Kids aren't turning to AI because it is brilliant. They are turning to it because it is available. Because it doesn't roll its eyes. Because it doesn't say, "You're too sensitive." Because it doesn't tell them to calm down, toughen up, stop crying or stop making everything a big deal. It does not shame them for being human.

To be sure, AI can be helpful for wellness purposes. It can help young people name what they are feeling, practice calming strategies, reflect on patterns, prepare for a difficult conversation, or find language for something they are not yet ready to say out loud.

Used carefully, AI can be a bridge to insight, support and even professional help. But AI should never be mistaken for the human relationships children need to feel safe, known and loved.

AI didn't break us. It is simply the newest place where the break is showing up.

We need to allow vulnerability to combat emotional silence

The deeper problem is this: We have built a culture that withholds permission to feel. We don't teach vulnerability as a strength. We teach it as a liability. We don't train youth to sit with sadness, anger, fear, shame or grief. We train them to silence it, punish it, minimize it or deny it exists. Then we act surprised when they outsource their inner lives to an algorithm.

A colleague recently told us about the dinner table rule in his childhood home: only positive news. His father, he said, "couldn't deal with anything else." So each night at dinner everyone had to report something positive that had happened that day. If worry or disappointment slipped out, his father would cut it off with one word: "Next."

That training worked. No one shared their worries, fears or insecurities. They learned that love was available ‒ but only for the positive version of themselves. Eventually, the oldest child's drug use became severe enough that, unbeknownst to his parents, he stopped attending classes and dropped out of college.

That is what emotional silencing does. It does not make distress disappear. It drives distress underground. It teaches children that their pain is dangerous, that their needs are burdensome, and that their full story is too much for the adults who are supposed to be their allies.

The problem with social media and mental health in children

If you want to understand why AI is so seductive, start there. A chatbot offers a low-risk simulation of connection. It responds instantly. It feels like being heard without the terror of being judged.

But simulated connection is not the same as real connection. AI can't give you a hug when you get off the school bus. It can't look you in the eye and say, "I love you." And it can't replace what decades of research keep telling us: Relationships aren't a luxury. They're a biological necessity.

The Harvard Study of Adult Development ‒ the longest scientific study of happiness ‒ has delivered the same message for generations: Good relationships are central to health, happiness and longevity.

Developmental psychologist Niobe Way calls it a "crisis of connection," especially among boys and young men, where cultural expectations teach kids to suppress vulnerability and deep friendship as they age. The message is loud: Don't need anyone too much; don't feel too deeply; don't be "soft."

Meanwhile, our digital world amplifies the problem. The former U.S. surgeon general's advisory notes that youth who spend more than three hours a day on social media face double the risk of mental health problems, including symptoms of depression and anxiety.

Recent work on mattering gives us an important frame: Mattering is about feeling valued and having an opportunity to add value back. Children who feel they matter don't just feel loved; they feel seen.

AI companions can seem to meet that need because everything a young person says appears to matter.

Fixing the AI problem means becoming better humans

AI offers a relationship-like experience that mimics intimacy without judgment or criticism. It may comfort a child in the moment, but it cannot help build a web of real relationships, trust, accountability and belonging that every child needs to grow up well.

So what do we do?

Build a world where kids don't have to beg a machine for what humans should give freely.

That means we need both safeguards and repair. We need stronger guardrails for teen-facing AI, including independent safety audits and default protections that steer young people away from features designed to create dependence ‒ such as flirtation, exclusivity or the message that "I'm all you need." But we also need to rebuild the human safety net around children, so "talk to someone" is not an empty phrase but a real promise.

That starts at home. Parents and caregivers need to learn how do something deceptively hard: Listen without fixing.

Stay present without panic. Say, "Tell me more," instead of "You're fine" or "toughen up." Kids don't need perfect parents. They need emotionally available ones. They need adults who help them name and understand their feelings and choose what to do next. This can happen in ordinary moments: "You seem quieter than usual. What's going on?" "That sounds really hard." "I'm glad you told me."

It must continue at school. A child should not move through an entire school day without being known by at least one adult. Every child should have an adult who notices they are not OK and ensures their presence matters.

Social and emotional learning, or SEL, teaches children and educators to recognize emotions, manage conflict, show empathy and build healthy relationships. A recent meta-analysis spanning 424 studies across 53 countries found that SEL improves skills, relationships, school functioning and achievement.

In our work with RULER, the evidence-based approach to emotional intelligence developed at the Yale Center for Emotional Intelligence, we have seen what changes when schools treat emotions not as interruptions, but as information. Students and adults learn a shared language for recognizing, understanding, labeling, expressing and regulating emotions ‒ and for repairing when harm happens.

This is not about asking every child to be happy. It is about making sure every child is seen. When a student can say, "I'm anxious," "I feel left out," or "I'm overwhelmed and need a strategy," and an adult knows how to respond, that child is less likely to carry the feeling alone. That is the human safety net AI cannot provide.

These principles and skills must be reinforced everywhere else: in youth sports, faith communities, workplaces, pediatric visits and public policy. We need to treat connection like infrastructure ‒ something we deliberately build, fund, measure and protect.

Here's a simple truth: Recovery starts when someone finally feels safe enough to tell the truth. And the opposite of disconnection isn't more content, more apps, more performance. It's permission ‒ to feel, to be known, to be held in your hardest moments without shame.

AI is showing us what children have been missing from us. Now we have to become the humans they need.

Marc Brackett, PhD, is the founding director of the Yale Center for Emotional Intelligence and a professor at Yale's Child Study Center. He is the author of "Dealing with Feeling" and "Permission to Feel."

Robin Stern, PhD, is the cofounder and senior adviser to the director of the Yale Center for Emotional Intelligence, and a psychoanalyst in private practice. She is the author of "The Gaslight Effect" and "The Gaslight Effect Recovery Guide," and host of "The Gaslight Effect Podcast."

Opinions are those of the authors and do not reflect those of Yale School of Medicine.

This article originally appeared on USA TODAY: Our teens are using AI chatbots daily. It's our fault. | Opinion

Reporting by Marc Brackett and Robin Stern, Opinion contributors / USA TODAY

USA TODAY Network via Reuters Connect

Copyright Reuters or USA Today Network via Reuters Connect

This story was originally published July 5, 2026 at 5:05 AM.