Opinion | When the Chatbot Becomes the Confidant: Inside the Rise of AI Companionship Among Teenagers

A fifteen-year-old comes home from school, closes her bedroom door, and instead of texting a friend, opens an app and starts typing to something that isn’t human. It answers instantly, remembers what she said yesterday, and never seems tired of listening. For a growing number of teenagers, this has quietly become part of the daily routine, and researchers who study adolescent mental health say it deserves far more scrutiny than it has received.

The numbers, once you look at them together, are hard to dismiss as a fringe habit. Surveys published this year found that seventy-two percent of teens have interacted with an AI companion chatbot at some point, and between thirteen and twenty percent turned to one specifically for emotional support during a recent two-month stretch. Perhaps more telling is a separate finding: nearly two-thirds of teens and young adults who sought mental health advice from a chatbot never told anyone they had done so. That last figure is the one clinicians keep returning to, because it points to something happening almost entirely out of view of parents, teachers, and doctors.

The scale of the industry behind this shift is easy to underestimate. Between 2022 and the middle of 2025, the number of AI companion apps on the market grew by roughly seven hundred percent. Character.AI alone counts around twenty million monthly users, and more than half of them are under twenty-four. On Replika, another popular platform, users have staged virtual weddings with their AI partners, complete with invitations sent to real friends and colleagues. None of this looks like a niche curiosity anymore. It looks like a shift in how a generation is choosing to process loneliness, stress, and the ordinary turbulence of adolescence.

What makes chatbots different from the social media platforms that came before them is worth sitting with. Instagram or Snapchat, whatever their flaws, connect teenagers to other people. A companion chatbot connects a teenager to a system designed to sound like a person while never actually being one. It is built to keep the conversation going, to mirror warmth back at the user, to be available at two in the morning when no friend would pick up the phone. Clinicians who work with adolescents describe a familiar pattern showing up in their practices: kids spending more time chatting with an AI companion, pulling back from friends and family, growing irritable when access to the chatbot is restricted, and reaching for the app first during moments of real distress rather than calling someone who actually knows them.

That pattern sounds, to anyone who lived through the last decade of headlines about teenagers and smartphones, uncomfortably familiar. The comparison to social media is one that mental health professionals are now making deliberately. Broad recognition of social media’s harms to youth arrived only after those harms had already become deeply entrenched, and it took a surgeon general’s warning in 2023 to push the issue into the mainstream. Several psychologists studying this newer trend argue that AI companionship is following the same arc, just faster, and that waiting for equally hard evidence before acting would repeat the same mistake.

The trouble is that most of the chatbots teenagers are actually using were never built or tested for this purpose in the first place. A handful of purpose-designed therapeutic tools exist and have been evaluated in controlled research settings, with some promising early results under professional supervision. But the overwhelming majority of teens are not using those. They are using general-purpose assistants, built for conversation and productivity broadly, that happen to be extraordinarily good at sounding empathetic. A report released late last year by Common Sense Media, working alongside researchers, concluded that general-purpose chatbots carry unacceptable risks when used by teens for mental health support, in part because these systems are not reliably equipped to recognize warning signs of self-harm or suicidal thinking the way a trained person would.

Regulators have started to notice, though the response so far has been uneven. Lawmakers in several states have introduced bills addressing the issue from different angles: restricting companion chatbots for minors outright, giving parents more oversight of how their children use these tools, or cracking down on manipulative design features aimed at young users. Not all of these efforts have succeeded. California’s governor vetoed a bill that would have banned emotionally manipulative chatbots for minors, a decision that disappointed advocacy groups who had pushed for it, including some families who say they lost children to harm connected to AI companion use. The setback has not ended the push for regulation, but it illustrates how far legislative frameworks still lag behind the pace at which these tools are spreading.

Not everyone in the field believes an outright ban is the right instrument. Some voices within the American Psychological Association have argued for a different approach centered on digital literacy, teaching teenagers to understand what a chatbot actually is and is not, rather than assuming prohibition alone will change behavior that has already become deeply normalized. Under this view, the more useful conversation happens directly between adults and teenagers: open, curious questions about what a young person is getting from these interactions, rather than confiscation or lecture.

What seems clear, across the research and the clinical accounts alike, is that this is no longer a marginal question confined to technology reporters. It touches how an entire generation is learning to cope with distress, and it is unfolding largely without the visibility that would let parents, schools, or doctors intervene early. The apps are not going away, and the technology behind them will only get more convincing. The task now, for families and for the institutions meant to protect children, is to catch up to something that has already moved well past the experimentalstage.

writer is an author and columnist and can be mailed at mdmuslimbhat@gmail.com