Psychiatry
Major Depressive Disorder
The Role of Artificial Intelligence and Digital Psychiatry in Major Depressive Disorder Care
Artificial intelligence (AI) is rapidly transforming some parts of psychiatry, helping with various administrative tasks. Chatbots that are being developed for cognitive behavioral therapy (CBT) may eventually be beneficial in patients with major depressive disorder (MDD) and other mental health concerns. Still, AI remains a subject of ongoing debate, demonstrating both real promise and risks, including patients using digital tools to obtain potentially inaccurate or incomplete mental health information that may lead to self-diagnosing or making treatment decisions without appropriate clinical evaluation.
AI has been described as the Fourth Industrial Revolution, and we are really only in the beginning stages. The early versions of AI models sometimes struggled, especially with clinical reasoning. These models performed adequately on multiple choice questions, but now they can answer complex clinical questions to the point where I can no longer give my students take-home examinations. The reasoning ability and knowledge base of generative AI models have gotten so good that they are probably going to change education in psychiatry and most aspects of what psychiatrists do.
Originally, the thought was that AI would be a useful tool in psychiatry, primarily for administrative tasks such as helping to write progress notes and history and physical notes. Now, as agentic AI is becoming more of a presence, it can also help with scheduling, preauthorizations, and medication refills, which is particularly helpful in patients with MDD who may need more frequent visits and complex treatment plans. In some ways, this can help make things easier for psychiatrists (eg, not having to use their time to perform many administrative tasks).
Agentic AI models that can be trained to execute standardized psychotherapies and to even be emotionally responsive are also now being developed. They are not perfect, but given the shortage of psychiatrists and the significant mental health needs of patients with MDD, could agentic AI therapy eventually fill a niche, particularly for people who have difficulty accessing psychiatrists or other mental health professionals?
These types of agentic AI models could help extend the reach of mental health care. For example, a chatbot could be trained to recognize language that suggests a suicide risk and could then reach out to a human being to intervene. There are also chatbots that are designed to deliver CBT, and they can really be quite good at it, performing tasks that include engaging in a phone conversation when patients are having trouble. Additionally, there are web-based CBT modules for anxiety, depression, and insomnia that have been demonstrated to be useful for patients.
There is no doubt that patients, particularly those with MDD, are now using generative AI models regularly to understand what they are dealing with from a mental health standpoint and to learn more about their medications. These models can look through the internet more effectively than a search engine and can synthesize information in ways that patients can understand. This is a really useful thing. However, there is also no doubt that patients are using AI in ways that it was not intended to be used. For example, patients sometimes self-diagnose based on incomplete information or come into the office saying, “I think I need this medication because ChatGPT [OpenAI] suggests that it would be potentially useful for me.” That is a risk.
I used to think that the field of psychiatry was relatively sheltered from AI compared with other medical disciplines because much of our work is based on interaction, empathy, and clinical judgment—things that people thought AI was not particularly good at. But now I am not so sure. As it turns out, patients do develop relationships with AI chatbots, although that sometimes leads them down a path that is not useful. On the other hand, some patients do not necessarily have the time to work, or interest in working, with a therapist (even with a telehealth option), but they are willing to go at their own pace with CBT modules online. Evidence suggests that these can be quite helpful to patients.
Banerjee S, Dunn P, Conard S, Ali A. Mental health applications of generative AI and large language modeling in the United States. Int J Environ Res Public Health. 2024;21(7):910. doi:10.3390/ijerph21070910
Frances A. Artificial intelligence and psychiatric diagnosis: a warning to the mental health community. World Psychiatry. 2026;25(1):148-149. doi:10.1002/wps.70015
Fuster-Casanovas A, Melero SM, Cabutto D, et al. Exploring digital health tools for depression management in primary health care: systematic review and meta-analysis. J Affect Disord. 2025;381:494-506. doi:10.1016/j.jad.2025.03.187
Kirk A, King CS, Gallop R, Doan BD, Simpson Rowe L, Randa LE. Evaluating a dual digital cognitive behavioral therapy and health and wellness coaching intervention for anxiety and depression: single-arm pilot study. JMIR Hum Factors. 2026;13:e92448. doi:10.2196/92448
Leung LB, Hovsepian S, Dyer KE, et al. What do primary care clinicians and patients think about internet-based computerized cognitive behavioral therapy for depression? A qualitative study from the Veterans Health Administration. Psychol Serv. 2025;22(3):440-448. doi:10.1037/ser0000943
Sharma D, Meshkat S, Perivolaris A, et al. Reimagining psychiatric care with agentic AI: promise, challenges, and a roadmap forward. NPJ Digit Med. 2026;9(1):252. doi:10.1038/s41746-026-02453-4
Stade EC, Tait ZM, Campione ST, Wiltsey Stirman S, Eichstaedt JC. Real-world use of large language models for mental health in 2024. NPJ Digit Med. 2026;9(1):630. doi:10.1038/s41746-026-02842-9
Torous J, Blease C. Generative artificial intelligence in mental health care: potential benefits and current challenges. World Psychiatry. 2024;23(1):1-2. doi:10.1002/wps.21148



