A JAMA Psychiatry trial finds an AI chatbot achieves outcomes comparable to human CBT therapy for mild-to-moderate depression across 3,000 patients.
A randomized controlled trial published in JAMA Psychiatry has found that an AI-powered therapeutic chatbot achieves clinical outcomes comparable to human-delivered cognitive behavioral therapy (CBT) for patients with mild-to-moderate depression. The study, conducted across 60 clinical sites in 8 countries, could help address the massive global shortage of mental health professionals.
The trial enrolled 3,000 adults diagnosed with mild-to-moderate major depressive disorder and randomized them to one of three groups: AI chatbot therapy, human therapist-delivered CBT, or a waitlist control. Over 12 weeks, participants in the AI group interacted with the chatbot through text and voice conversations that delivered structured CBT modules including cognitive restructuring, behavioral activation, and problem-solving therapy.
At the end of treatment, the AI chatbot group showed a mean reduction of 8.4 points on the PHQ-9 depression scale, compared to 9.1 points in the human therapy group and 2.8 points in the control group. The difference between AI and human therapy was not statistically significant, meeting the pre-specified non-inferiority margin.
Critically, the AI chatbot was available 24 hours a day, 7 days a week, and participants used it an average of 4.2 times per week, compared to the standard once-weekly human therapy sessions. The convenience and accessibility of the chatbot resulted in higher engagement rates and lower dropout rates: 89% of AI group participants completed the 12-week program, versus 71% in the human therapy group.
The chatbot used a large language model fine-tuned on therapeutic conversation data from licensed psychologists. It incorporated safety protocols including real-time risk assessment, automatic escalation to human crisis counselors for suicidal ideation, and regular clinical oversight by supervising psychologists who reviewed anonymized conversation summaries.
Dr. John Torous of Beth Israel Deaconess Medical Center, who led the study, emphasized that the AI chatbot is not intended to replace human therapists but to extend the reach of mental health services. An estimated 75% of people with depression worldwide receive no treatment, primarily due to insufficient numbers of trained professionals. AI-assisted therapy could provide evidence-based support to millions who would otherwise go untreated.
The researchers noted important limitations. The chatbot was less effective for severe depression and complex cases involving comorbid conditions, where the nuance and flexibility of human therapists proved superior. The recommended approach is a stepped-care model in which AI chatbots serve as the first line of treatment for mild-to-moderate cases, with human therapists reserved for complex presentations and cases that do not respond to initial AI-assisted treatment.
Several healthcare systems are already piloting integration of AI therapeutic tools. The UK's National Health Service has launched a program offering AI chatbot therapy as part of its Improving Access to Psychological Therapies service, and Australia's Beyond Blue has incorporated similar technology into its digital mental health platform.