Dartmouth Center for Technology and Behavioral Health (CTBH)

Individuals with major depressive disorder (MDD), generalized anxiety disorder (GAD), and clinically high-risk feeding and eating disorders (CHR-FED) face significant barriers to accessing timely mental health treatment. There is a critical need for scalable, accessible interventions that can bridge the gap between demand and the availability of trained clinicians.

Researchers deployed Therabot, a generative AI–powered text-based chatbot, in a randomized controlled trial (N=210) across three diagnostic groups. Participants in the intervention arm were prompted daily for four weeks to engage with Therabot, then retained access for an additional four weeks post-intervention without daily prompts.

Therabot use was associated with statistically significant reductions in clinical symptoms across all three diagnostic groups (MDD, GAD, CHR-FED) at both 4-week and 8-week timepoints compared to waitlist controls. Participants reported forming a therapeutic alliance comparable to traditional outpatient psychotherapy (mean WAI score = 3.59), and reported high satisfaction and willingness to use it independently. Safety interventions were required 15 times for suicidal ideation and 13 times for inappropriate responses.

• Generative AI chatbots can achieve clinically meaningful symptom reductions across multiple mental health diagnoses in a short-term trial, suggesting viability as a scalable adjunct to care.
• User engagement was high (95% interaction rate, avg. 260 messages over 24 days), indicating strong acceptability for a fully digital intervention.
• Safety monitoring remains essential: human staff intervention was required for suicidal ideation and inappropriate AI responses, underscoring that Gen-AI chatbots cannot yet operate fully autonomously in clinical settings.