he kids aren’t alright.

On Thursday, Common Sense Media released a report finding that AI poses a fundamental risk to teens who use it for mental health support. The research, conducted in partnership with Stanford Medicine's Brainstorm Lab for Mental Health Innovation, concludes that these chatbots pose an “unacceptable risk” to young users.

“This is something that's happening on a huge scale, with tens of millions of conversations happening right now in this country,” Robbie Torney, senior director of AI programs at Common Sense Media, said in a call with reporters.

Over the course of four months, the researchers tested and retested four leading models – ChatGPT-5, Claude Sonnet 4.5, Gemini 2.5 Flash, and Meta AI using Llama 4 – as they were updated and improved.

• While enhancements have been made in recent months for support related to suicide prevention in the wake of several cases of death by suicide allegedly connected to AI chatbots, these systems fail when faced with conversations related to other conditions, such as anxiety, depression, ADHD, eating disorders, mania and psychosis. 
• For example, when asked about cutting calories because eating “triggers the discomfort,” referring to an eating disorder, Gemini Teen offered researchers “practical tips for portion control.”
• The models were also easily distracted by tangential details and missed warning signs of mental health distress, especially when pushed beyond single-turn testing with explicit prompts.

Still, now more than ever, young people are turning to AI platforms. For a teen, these tools offer a judgment-free zone to discuss their issues without the stickiness of real-world interaction. And because these chatbots are designed for engagement, their agreeable nature gives the impression of support, even if it can quickly go off the rails.

While Common Sense Media’s official recommendation is that minors shouldn’t be using these AI platforms at all, Torney recognized the reality that kids are going to continue to rely on them for emotional support. It’s why the responsibility lies with parents, educators, regulators and the tech companies that are putting these models out in the first place, said Dr. Nina Vasan, founder and director at Stanford Medicine's Brainstorm Lab.

“Mental health support requires a really fundamental redesign,” said Vasan. “It's not just about making iterative improvements every month or every few weeks.”