A Letter about “Randomized Trial of a Generative AI Chatbot for Mental Health Treatment”

A recently published randomized controlled trial of a generative AI chatbot for mental health treatment, "Therabot," requires a response. The authors report reductions in symptoms of depression, anxiety, and feeding and eating disorders, as well as the formation of a therapeutic alliance comparable to outpatient psychotherapy. We identify three methodological limitations that undermine confidence in these conclusions: the use of a wait-list control, the lack of independent evaluation, and the misapplication of a measure developed for human therapeutic relationships. Although we recognize the potential of digital therapeutics and interactive self-help tools such as Therabot to expand access to mental health treatment, we suggest that more robust research is needed before the study's conclusions can be supported, both to prevent premature commercialization and to minimize clinical and ethical risks.