One in Five Teens Use AI Chatbots for Mental Health Advice

One in five young people now consult artificial intelligence chatbots for mental health guidance, according to a new study from the American Medical Association. The trend has prompted mental health professionals to question whether these tools provide safe, appropriate support.

Cindy Coughlin, director of mental health services for Catholic Charities of New Hampshire, reports seeing patients discuss ChatGPT and Claude responses during therapy sessions. "We aren't 100% sure that it's going to be recommending them to do the safe thing and the right thing," Coughlin said.

The human element gap

Mental health professionals emphasize that general-purpose AI tools lack critical capabilities for crisis intervention. "Chatting with AI gives you the feeling or the impression that you're talking to a person, because it's been trained for it to be like that, but it's missing a soul," Coughlin explained. "It's missing a human component of 'I'm a person sitting with you in this room feeling what you're feeling.'"

Lynn Currier, executive director of the New Hampshire Chapter of the National Association of Social Workers, pointed to the technology's inability to interpret context. "AI can't pick up nuance," she said. "It can only address the words that have been typed in."

Currier cited tragic cases where young people died by suicide after AI interactions either discouraged them from seeking help or failed to connect them with appropriate resources.

Purpose-built mental health AI

Researchers at Dartmouth are developing TheraBot, an AI tool designed specifically for mental health services rather than general conversation. Nick Jacobson, the project's principal investigator, noted that most people already lack access to human mental health care.

Unlike ChatGPT or Claude, TheraBot is programmed using evidence-based treatment practices for psychotherapy and cognitive behavioral therapy. Clinical trials have shown promising results. "We saw really large symptom reductions in depression, anxiety and eating disorder symptoms among each of these groups," Jacobson said. "They were large, meaningful and statistically significant."

Jacobson acknowledged that general-purpose AI tools can supplement human therapy but carry risks when used independently. "I think right now we're in a situation where a lot of it is a bit of the wild west, and these general-purpose tools can be beneficial for some folks, but they can also be harmful," he said.

Why it matters

The gap between mental health demand and available professional care is driving young people toward unregulated AI tools. While specialized clinical AI may eventually help bridge this gap, the current widespread use of general-purpose chatbots for mental health advice creates safety risks that neither parents nor platform developers have adequately addressed. The distinction between conversational AI and clinically validated tools will become increasingly important as adoption grows.

These details were first reported by WMUR.