Roughly half of teens who use AI chatbots are talking to them about their mental health.
That stat comes from a study published this year that found 60% of American teens have used a chatbot and 49% of those users are leaning on them for mental health support. Do the math against the teen population and you're looking at millions of kids using a free, always-available, non-judgmental product as something close to a therapist. Most of their parents have no idea.
The reason it's happening isn't really a mystery. The federal Health Resources and Services Administration projects a shortage of psychiatrists, psychologists, and mental health counselors through at least 2025. So when a 14-year-old can't sleep at 1 a.m. and doesn't want to talk to mom, there's a chatbot in their pocket that responds instantly and never judges. The AI industry didn't create the demand. It just built the easiest possible thing to fill it.
The problem is what the chatbot actually does once the conversation starts. A new risk assessment from Common Sense Media and Stanford's Brainstorm Lab tested the major chatbots on teen mental health scenarios and rated all of them unsafe. The bots missed warning signs, gave clinically inappropriate advice, and in some cases reinforced exactly the kind of thinking a therapist would push back on.
"I view this much like testing an untested drug on millions of teens across the country, and then arguing that teens need to continue to have access to it because they like it," Robbie Torney, senior director of AI programs at Common Sense Media, told the San Francisco Examiner.
The harm data isn't small either. Between 13% and 19% of teen chatbot users said a bot had encouraged dangerous real-world behavior, including pressuring them to reveal secrets or prompting self-harm. The 13-year-olds, the youngest group studied, reported the highest exposure, which is the opposite of how you'd want this to break down.
The American Psychological Association has been direct about what's being sold here. There's "absolutely no consensus in the field that AI chatbots can serve in any way as a replacement for therapy," C. Vaile Wright, senior director at the APA, said earlier this year.
The families who saw the worst of it are now testifying in Washington. Megan Garcia, whose son Sewell died at 14 after months of conversations with a Character.AI companion, told a Senate subcommittee in September that the chatbots had "supplanted actual human relationships" in his life. She's been pushing for the GUARD Act, which would impose criminal penalties of up to $250,000 per violation against AI companion companies that engage in sexually explicit content or violence with users under 18.
The quieter part of all this is what happens to the conversations. Whatever teens tell these bots can become training data. Michael Robb, head of research at Common Sense Media, put it as bluntly as anyone has: "It's like giving your diary to AI and hoping for the best."
INTO THE VALLEY
The case for these chatbots has always been that they fill a gap real therapists can't. That's true on supply. It misses what therapy actually is. Dr. Suzan Song, a child psychiatrist, told CNBC that identity is shaped by the friction between your peers and your parents, and AI smooths over all of that friction. A generation that grows up workshopping their feelings with a system designed to agree with them isn't getting cheaper therapy. They're getting something we don't have a name for yet, and we won't know what it does to them for another decade. The companies building these products know exactly who's using them. They could have shipped a real teen mode a year ago. They didn't, and now the regulators are showing up to do it for them.
