Artificial Intelligence (AI) has prompted a lot of concerns, but there are also ways the simulation of human intelligence by machines has made our lives better — from improving customer service to providing quick access to information.

Artificial intelligence may also make therapy more accessible. An AI chatbot called “Therabot” significantly improved the mental health of people with major depressive disorder (MDD), generalized anxiety disorders (GAD) and eating disorders, according to a Dartmouth University study.

The relationships between the AI “therapists” and the participants were surprisingly close.

Generative AI, also known as GenAI or GAI, are machine learning models that can produce new content such as text, images and videos. GenAI chatbots go one step further by producing human-like content.

For this study, GenAI Therabot was fine-tuned on mental health conversations created by mental health experts that included a clinical psychologist and board-certified psychiatrist. In addition, the conversations were peer-reviewed specifically using cognitive behavior therapy (CBT) models. CBT is a structured form of talk therapy.

The trial included 106 people from across the United States who were diagnosed with MDD, GAD or an eating disorder. The participants interacted with Therabot through a smartphone app by typing out their responses to prompts about how they were feeling or initiating conversations when they needed to talk.

The results were promising: After the 8-week long study:

  • People diagnosed with depression experienced a 51 percent average reduction in symptoms, leading to clinically significant improvement in mood and overall well-being.
  • Participants with generalized anxiety reported an average reduction in symptoms of 31 percent with many shifting from moderate to mild anxiety, or from mild anxiety to below the clinical threshold for an anxiety diagnosis.
  • Those at risk for eating disorders are traditionally challenging to treat. But the Therabot users showed a 19 percent average reduction in their unrealistic concerns about body image and weight.

These numbers point to the effectiveness of using Therabot to treat mental illness.

“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,” Nicholas Jacobson, the study's senior author and an associate professor of biomedical data science and psychiatry at Dartmouth's Geisel School of Medicine, said in a press release.

Therabot was fine-tuned on mental health conversations created by mental health experts that included a clinical psychologist and board-certified psychiatrist.

The findings come at a time when there is a growing shortage of mental health professionals. For instance, for every available provider in the country, there's an average of 1,600 patients with depression or anxiety alone. “There is no replacement for in-person care, but there are nowhere near enough providers to go around,” Jacobson pointed out.

The relationships between the AI “therapists” and the participants were surprisingly close. The research team didn't expect that they would almost treat the software like a friend. One indication of this bond was that the participants not only provided detailed responses to Therabot's prompts — they frequently initiated conversations. Interactions with the software also showed upticks at times usually associated with unwellness — such as in the middle of the night.

When people initiated a conversation with the app, Therabot answered with natural open-ended text dialog based on an original training set the researchers developed from current, evidence-based best practices for psychotherapy and cognitive behavioral therapy.

That said, the study's authors are not suggesting the psychology community abandon in-person therapy and rush to develop more AI bot counseling programs. “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,” said Michael Heinz, the study's first author and an assistant professor of psychiatry at the Center for Psychology and Behavioral Health, Dartmouth and Geisel School of Medicine, as well as attending psychiatrist at Dartmouth Hitchcock Medical Center.

If you would like to speak with a mental health professional, start by contacting your primary care physician, trusted friends or family, or exploring resources like the National Alliance on Mental Illness (NAMI), Mental Health America, or the Psychology Today “Find a Therapist” tool.

You can find out more about Therabot here.

The study is published in the New England Journal of Medicine.