Researchers at Dartmouth College have published the first randomized controlled trial to evaluate the effectiveness of a generative AI chatbot, Therabot, for treating mental health symptoms. The study enrolled 210 participants with symptoms of major depressive disorder (MDD), generalized anxiety disorder (GAD), and clinically high-risk feeding and eating disorders (CHR-FED). Participants were randomly assigned to a waitlist control or a 4-week intervention using Therabot, and researchers evaluated participants symptoms at baseline, at the conclusion of the study (4 weeks), and at follow up (8 weeks).
Results showed that participants assigned to work with Therabot had clinically significant reductions in symptoms of MDD, GAD, and CHR-FED—both immediately following the intervention and at the follow-up—compared with participants in the waitlist condition. At the 8-week follow up, Therabot users reported a 51% reduction in MDD symptoms, a 31% reduction in GAD symptoms, and a 19% reduction in CHR-FED symptoms compared with their baseline symptom reports. Additionally, participants engaged with Therabot frequently—sending an average of 260 messages and spending an average of over 6 hours messaging with Therabot throughout the course of the study. Participants also rated their therapeutic alliance with the chatbot similarly to that of individuals engaged in outpatient psychotherapy and were generally satisfied with their therapeutic experience.
Researchers were optimistic when discussing these findingsopens in new window, explaining that the improvements participants reported are comparable to the rates of improvement seen in outpatient therapy. They also assert that AI interventions have a role to play in addressing the current mental health needs of the public, given the ability of AI tools to provide real-time support to individuals outside of the traditional therapy room. However, they also cautioned that GenAI chatbots are not ready to function autonomously (e.g., without human oversight) in mental health care and that more research across broader populations is needed. Still, this research is an important step in beginning to establish the safety and effectiveness of GenAI chatbots for treating mental health conditions. For a deeper dive into this research, read the full articleopens in new window.
Additionally, in an effort to advance AI research, Wellcome Trust, a UK-based charitable foundation, has announced a new funding opportunity for research teams investigating how generative AI can be used to improve the treatment of anxiety, depression, and psychosis. The program has two phases. First, teams would be accepted into an accelerator program that will support the development of pilot studies. Then research teams who participated in the accelerator would be eligible to apply for funding. Applicants could be awarded up to £3 million (~3.9 million USD); the deadline for accelerator applications is May 28, 2025. Please see the full description of the opportunity and requirementsopens in new window.
