Michael V. Heinz, Daniel M. Mackin, Brianna M. Trudeau, Sukanya Bhattacharya, Yinzhou Wang, Haley A. Banta, Abi D. Jewett, Abigail J. Salzhauer et al.
A randomized controlled trial of the expert-fine-tuned generative AI chatbot Therabot demonstrated symptom improvement in depression, anxiety, and eating disorder risk groups.
There is a need for methods to provide personalized mental health treatment at scale while addressing low user engagement and retention issues common in digital therapeutics.
210 adult participants were randomly assigned to a 4-week Therabot intervention or waitlist control. Participants were stratified into three groups based on clinically significant symptoms of MDD, GAD, or CHR-FED. Symptom changes were assessed using cumulative-link mixed models, and Cohen's d effect sizes were calculated.
After 4 weeks, all strata showed significant symptom reduction compared to controls (moderate to large effect sizes). Effects were maintained at 8-week follow-up. High user engagement and therapeutic alliance scores suggest that generative AI chatbots may be effective in real clinical settings.