In the first clinical trial of its kind, Dartmouth’s AI-powered therapy chatbot Therabot significantly reduced symptoms of depression, anxiety and eating disorders—offering real-time, evidence-based support comparable to human therapists. (Source: Image by RR)

Dartmouth Researchers Validate Therabot’s Safety and Efficacy in Real-World Use

Dartmouth researchers have conducted the first clinical trial using a generative AI-powered therapy chatbot, known as Therabot, to treat individuals with mental health disorders such as depression, anxiety and eating disorders. The trial included 106 participants across the U.S. and revealed significant improvements in symptoms—depression symptoms were reduced by 51%, anxiety by 31%, and eating disorder concerns by 19%. Therabot, accessed via a smartphone app, engaged with users through natural language conversation and prompted responses based on best practices from psychotherapy and cognitive behavioral therapy.

Users interacting with Therabot developed a strong sense of trust and therapeutic alliance with the AI, comparable to in-person therapy. Many participants initiated conversations themselves, often during vulnerable moments such as late at night, and reported feeling emotionally connected to the chatbot. As noted in an article in home.dartmouth.edu, researchers found that after eight weeks of use, the improvements in participants’ mental health were on par with gold-standard outpatient therapy. Importantly, 75% of those in the trial were not receiving other treatment, suggesting Therabot could be a critical support option for underserved populations.

The technology behind Therabot includes AI-generated responses trained on evidence-based therapeutic frameworks and is capable of identifying high-risk content such as suicidal ideation, offering crisis resources as needed. The app used a stamping-like method to personalize interactions and adapt based on user input over time. While the results are promising, researchers emphasized the importance of close clinician oversight, noting that although safety risks were rare during the trial, generative AI’s open-ended nature requires robust safeguards and rapid-response capabilities in clinical settings.

Therabot represents a broader shift in mental healthcare toward scalable, always-available AI companions capable of supplementing traditional therapy. The success of this trial supports the idea that generative AI can provide meaningful mental health support, especially where access to human professionals is limited. However, the researchers stress that careful, ethical deployment and evaluation must continue. With the increasing interest in AI therapy tools following the rise of ChatGPT, the Dartmouth team sees their study as a model for combining technological innovation with rigorous mental-health oversight.

read more at home.dartmouth.edu