In 2025, one of the world's most prestigious medical journals — NEJM AI — published results from the first-ever randomized controlled trial of a generative AI therapist. The headline finding: a 51% reduction in depressive symptoms. This isn't marketing copy — it's peer-reviewed science.

How it worked

The study was called Therabot. It enrolled 210 adults diagnosed with major depressive disorder, generalized anxiety disorder, or eating disorders. Half the participants spent four weeks talking with the AI therapist; the other half were placed on a standard waitlist. On average, participants logged over six hours of interaction with the system across the trial period.

The results were stronger than most expected. Depressive symptoms dropped by an average of 51%, anxiety by 31%, and eating disorder symptoms by 19%. Participants also rated the therapeutic alliance — their sense of being understood and trusting the interaction — at levels comparable to those reported with human therapists.

What came before Therabot

Therabot wasn't the first attempt to build an AI therapist. Woebot, launched back in 2017, showed a modest positive effect on depression in college students in a small-scale study. Wysa has handled more than 400 million conversations across 65 countries. Apps like Tess and Youper have also shown some benefit — but none had been put through a trial with this level of methodological rigor.

What to keep in mind

Critics of the Therabot study raise fair points about its limitations. The control group simply waited — they received no active treatment. That design makes it impossible to say whether the AI performed better or worse than a human therapist. There's also the question of oversight: every conversation in the trial was monitored in real time by a team of more than 100 researchers. That kind of safety net would be difficult to replicate in a real-world deployment.

Still, the significance of this study is hard to overstate. For the first time, a generative AI has completed a clinical trial and produced measurable outcomes comparable to conventional therapy. That doesn't mean AI is about to replace therapists. But it is the first serious evidence that AI could become a genuinely useful tool — especially in places where a trained clinician isn't available, or where the waitlist stretches on for months.