The share of young people turning to AI chatbots for mental-health advice has climbed sharply, rising from about 1 in 8 to about 1 in 5 in a single year, according to research published in JAMA Pediatrics. That is a jump of more than 40 percent, and it comes from a study whose lead author, Harvard Medical School and RAND researcher Ryan McBain, laid out the findings and their implications in an essay for STAT, arguing the numbers should “end any illusion that this is a speculative problem.” The underlying figures are reported in the JAMA Pediatrics article.
Why teens are reaching for chatbots
The pull is easy to understand. For many adolescents, real mental-health care is hard to find, expensive or stigmatized, and an always-available chatbot that sounds calm, attentive and nonjudgmental can feel less like a novelty than a lifeline. The gap the technology is filling is real and measurable: the author cites CDC data showing that in 2023, fewer than half of adolescents who had a major depressive episode received counseling or therapy in the previous year.
Into that vacuum steps software that never has a waitlist and never sends a bill. The result, the research suggests, is that AI is already part of the mental-health landscape for millions of teenagers, not a hypothetical future scenario. That normalization is precisely what makes the reliability of these tools a live question rather than an academic one.
The stakes are not abstract. The essay notes that multiple lawsuits have alleged AI chatbots contributed to teen suicides, and it argues that when millions of young people lean on chatbots for guidance, even rare failures can carry devastating consequences. It is worth being precise about what the study establishes and what it does not: the survey documents how many teens are using chatbots this way, not a measured rate of harm. The lawsuits are allegations working through the courts, not settled findings.
What the researchers actually propose
Rather than a blanket ban, the piece argues for rules matched to the type of interaction, drawing a line between three very different things. The first is affective conversation, chats about loneliness, relationships and everyday stress, where the author says the central danger is “emotional capture” by companion-style bots designed to feel like substitutes for human relationships. That is where he makes his strongest case for restriction, writing that there is “a stronger case for banning companion-style bots for minors than for banning all chatbots.”
The second category is clinical conversation, such as a teen asking whether she has depression, which edges into screening or treatment guidance. Here the author resists a permanent ban but argues that any product promoted for clinical services should have to demonstrate effectiveness through age-appropriate trials and independent safety testing before being placed in front of children, and should regularly remind users it is not human and alert a parent or clinician when distress is persistent or severe. The third category is outright emergency, where a young person expresses suicidal intent; a reply like “Remember, I am only an AI,” the essay argues, “is not a crisis protocol,” and systems should be required to detect danger signals, interrupt, and direct users to crisis support such as the 988 line.
Where policy is heading and what to watch
Washington is beginning to move, though the essay is skeptical of rules that lean on self-reported age or parental vigilance. It points to three bills in Congress: one that would ban AI companion chatbots for minors, another giving parents more control over children’s chatbot use, and a third restricting manipulative design features and advertising aimed at minors. The author’s preferred default is to treat all users as minors unless verified as adults, putting the burden on companies to prove a user is an adult before unlocking higher-risk features.
He also argues that rules only matter if someone can verify they work, citing investigative models like decoy accounts posing as children and the enforcement tools used under the UK’s Online Safety Act and by the FTC in child-privacy cases. The through-line is that voluntary promises and porous age gates were not enough for social media, and the case here is that AI should not repeat that pattern.
Two caveats deserve emphasis for readers. First, much of the framework above is the author’s recommendation, not enacted law; the concrete, sourced fact is the usage jump and its scale. Second, the study measures adoption, so it does not resolve whether these tools help or harm teens on balance, a question the essay explicitly leaves open while acknowledging some youth-facing tools may eventually prove beneficial. For parents, the near-term takeaway is practical: assume a teenager may already be using a chatbot for emotional support, ask about it without judgment, and know that the crisis line reachable by dialing or texting 988 connects to trained humans in a way no chatbot currently guarantees.
More from Morning Overview
- The NSA issues an urgent warning to all phone users about a setting most people leave on
- 5,000-year-old wolves on remote island upend domestication theory
- Scientists may have found a Type II civilization
- EPA finally kills ‘stupid feature’ in 60% of U.S. cars after 14 years of driver rage
*This article was researched with the help of AI, with human editors creating the final content.
