AI Therapist Matches Human Care in Groundbreaking Mental Health Trial

People struggling with depression who used an artificial intelligence therapy app for just four weeks experienced a 51% reduction in symptoms—matching results typically seen with human therapists, according to the first clinical trial of its kind.

The study, published March 27 in the New England Journal of Medicine AI, provides compelling evidence that AI could help address America’s critical shortage of mental health providers while delivering clinically meaningful care.

Researchers at Dartmouth College developed and tested “Therabot,” a smartphone-based chatbot designed to provide evidence-based therapeutic support through natural text conversations. The system was trained on current best practices for psychotherapy and cognitive behavioral therapy.

“The improvements in symptoms we observed were comparable to what is reported for traditional outpatient therapy, suggesting this AI-assisted approach may offer clinically meaningful benefits,” said Nicholas Jacobson, the study’s senior author and an associate professor of biomedical data science and psychiatry in Dartmouth’s Geisel School of Medicine.

The trial enrolled 106 participants diagnosed with major depressive disorder, generalized anxiety disorder, or eating disorders. They interacted with Therabot through a smartphone app, responding to prompts or initiating conversations whenever they needed support. Another 104 participants with similar diagnoses formed a control group.

Results showed anxiety sufferers experienced a 31% average reduction in symptoms, with many moving from moderate to mild anxiety classifications. Even participants with eating disorders—traditionally more challenging to treat—showed a 19% reduction in body image and weight concerns, significantly outpacing the control group.

“Our results are comparable to what we would see for people with access to gold-standard cognitive therapy with outpatient providers,” Jacobson explained. “We’re talking about potentially giving people the equivalent of the best treatment you can get in the care system over shorter periods of time.”

The availability gap in mental health care remains staggering. For every available provider in the United States, approximately 1,600 patients are suffering from depression or anxiety alone.

“There is no replacement for in-person care, but there are nowhere near enough providers to go around,” Jacobson noted. “We would like to see generative AI help provide mental health support to the huge number of people outside the in-person care system.”

Perhaps most surprising was how participants formed emotional connections with the digital therapist. Users reported levels of “therapeutic alliance”—the crucial bond of trust between patient and therapist—comparable to what patients typically experience with human providers.

“We did not expect that people would almost treat the software like a friend,” Jacobson said. “My sense is that people also felt comfortable talking to a bot because it won’t judge them.”

Many participants initiated conversations unprompted, particularly during vulnerable periods like late at night when human therapists would be unavailable.

Study participants engaged with Therabot for an average of six hours throughout the trial—equivalent to about eight therapy sessions. Nearly 75% were not receiving any other form of treatment during this period.

Despite the promising results, researchers emphasize that AI therapy still requires significant human oversight. The study team carefully monitored interactions and stood ready to intervene if concerning content emerged.

“While these results are very promising, no generative AI agent is ready to operate fully autonomously in mental health where there is a very wide range of high-risk scenarios it might encounter,” cautioned Michael Heinz, the study’s first author and an assistant professor of psychiatry at Dartmouth.

If Therabot detected potential self-harm content, it was programmed to immediately provide crisis resource options, including prompts to call 911 or suicide prevention hotlines.

The Dartmouth team has been developing Therabot since 2019, well before the recent explosion of generative AI tools like ChatGPT. Their cautious, methodical approach stands in contrast to the rush of mental health apps now incorporating AI without similar clinical validation.

“There are a lot of folks rushing into this space since the release of ChatGPT, and it’s easy to put out a proof of concept that looks great at first glance, but the safety and efficacy is not well established,” Jacobson warned. “This is one of those cases where diligent oversight is needed, and providing that really sets us apart in this space.”

As research continues, the team envisions AI therapy not as a replacement for human providers but as a complementary tool that could dramatically expand access to mental health care for millions currently unable to receive help.