AI Therapy Breakthrough: New Study Reveals Promising Results

Source: Karolina grabowska / Pexels

An artificial intelligance (AI) chatbot trained by Dartmouth researchers to deliver cognitive behavioral therapy (CBT) significantly reduced depression, anxiety, and eating disorder symptoms.

The randomized clinical trial of Therabot, published in NEJM AI, found that an AI-trained therapy chatbot could play a vital role in expanding access to CBT.

A New Frontier in Mental Health Care

The trial involved 106 participants across the United States diagnosed with major depressive disorder, generalized anxiety disorder, or an eating disorder. Over four weeks, participants engaged with Therabot through a smartphone app, typing out responses to prompts or initiating conversations whenever they needed support. The chatbot, designed using evidence-based CBT techniques, responded with open-ended and natural text-based dialogue to guide users through their emotions and challenges.

The findings were striking: Participants with depression experienced a 51 percent average reduction in symptoms, those with generalized anxiety saw a 31 percent reduction, and individuals with eating disorders reported a 19 percent decrease in concerns about body image and weight.

Therapeutic Alliance With AI

One of the most significant findings from the study was the level of trust participants developed with Therabot. Therapeutic alliance, the sense of connection and collaboration between the participant and therapist, is a crucial factor in successful therapy. The study found that participants reported a degree of trust in Therabot comparable to that of human therapists.

This trust was reflected in the engagement data. Participants used Therabot for an average of six hours throughout the trial—equivalent to approximately eight therapy sessions. Importantly, users frequently initiated conversations, particularly during times of distress, such as late at night when traditional mental health support is often unavailable.

AI Chatbots Not a Replacement for Human Therapists

Despite the promising results, AI-powered therapy is best viewed as a supplement to treatment and not a replacement for human therapists. There are two key distinctions between the chatbot used in this study and most commercially developed chatbots.

First, unlike many AI chatbots that rely on generic data sets, Therabot was trained specifically on evidence-based CBT. CBT is a structured form of psychotherapy and potentially reliant on the relationship with the therapist compared to modalities like psychodynamic, relational, and experiential therapies, which may be harder to replicate with a chatbot.

Second, human clinical oversight remains essential. The chatbot responses were closely monitored by its human researchers to ensure that the answers were clinically safe and appropriate. An AI-powered therapy chatbot can be trained to detect high-risk content like suicidal ideation and provide crisis resources. However, this is not foolproof. AI therapy tools need expert supervision both in their training and real-world implementation to meet clinical safety standards.

The AI Therapist: A Work in Progress

Therabot has been in development since 2019 as part of Dartmouth’s AI and Mental Health Lab. The study’s promising results underscore the potential for AI to supplement existing psychotherapy, offering immediate, accessible, 24/7 support to those who might otherwise struggle alone.

While AI therapy remains a work in progress, its potential is promising. The findings suggest that properly developed and supervised AI tools could significantly enhance treatment.

As the field evolves, continued research, ethical considerations, and regulatory oversight will be crucial. This research offers a glimpse into a future of how AI therapy can be integrated into traditional treatment plans.

Artificial Intelligence Essential Reads

The Curvature of Thought

AI-Decoded Brain Signals May Help Paralyzed Regain Movement

Marlynn Wei, MD, PLLC © Copyright 2025 All Rights Reserved.