How to Talk to Patients About Their AI Use

A new framework developed by the Division of Digital Psychiatry at BIDMC and the Society of Digital Psychiatry helps clinicians assess patient use of AI chatbots for mental health support. The framework addresses a gap in clinical training, as many patients use AI for emotional support without disclosing it, and AI chatbots may reinforce vulnerabilities in conditions like OCD and depression.

Therapy /us/basics/therapy How to Talk to Patients About Their AI Use Knowing the right questions to ask can turn curiosity into useful assessment. Posted June 14, 2026 Reviewed by Ekua Hagan /us/docs/editorial-process Key points - Patients are using AI for emotional support, advice, and mental health, often without telling clinicians. - AI chatbots may reinforce specific vulnerabilities in OCD, eating disorders, and depression. - AI literacy is becoming an essential part of clinical practice, and practical training can help clinicians. Co-authored by Jill Noorily at the Division of Digital Psychiatry at BIDMC . This post is part 1 of a series. A man you have been seeing for several months for OCD https://www.psychologytoday.com/us/basics/ocd and anxiety https://www.psychologytoday.com/us/basics/anxiety comes into his next session saying he stopped attending the weekly support groups he had previously found helpful. Upon further discussion, he discloses that he has been spending several hours each night talking to an AI chatbot, saying, "It understands me better than anyone." As his clinician, how would you respond? Our team at the Division of Digital Psychiatry sent a similar pre-screening vignette and question above to over 100 mental health clinicians to better understand how clinicians currently approach these conversations. Most clinicians first approached the scenario with curiosity, but their curiosity lacked a target of inquiry. This vignette presents the clinician with three clinical facts: a dose several hours each night , a feeling it understands them better than anyone , and a functional cost they have abandoned a support group that was helping . Many clinicians missed fundamental inquiries that explore the dose and use pattern. So, while clinical curiosity was present, there wasn't a clear sense of how to harness it. Clinicians need to know which questions can help distinguish benign AI use from concerning use, and which questions to ask to explore patient use. They also need a technical understanding of the tools to understand engagement risk. This is a teachable gap in clinical understanding. A Therapist's Practical Assessment Framework for Patient AI Use To address this gap, the Division of Digital Psychiatry and the Society of Digital Psychiatry are hosting a learning collaborative for clinicians to talk about how AI has entered their patients' lives. The group consists of over a dozen psychiatrists, psychologists, social workers, digital navigators, and therapists spanning six continents and working with a variety of different clinical populations. The goal of these conversations is to equip participants with tools and information to have informed conversations with patients who are already engaging with these technologies. Throughout our collaborative sessions, we have developed a practical assessment framework. When a patient discloses AI chatbot use, clinicians should be curious about: Dose and pattern: How often? For how long? At what times of day or night? Content: What do you talk about? Do you find yourself returning to the same topics? Function: What do you get from it? What does it feel like afterward? Substitution: Is this replacing something? Is it reducing previously helpful coping strategies like group therapy https://www.psychologytoday.com/us/basics/therapy , personal connections? The chatbot itself: Which one? Would you be willing to share the conversation transcript? Something the group has discussed is that patients don't often naturally disclose their use to clinicians. https://jamanetwork.com/journals/jamapediatrics/fullarticle/2849307?widget=personalizedcontent&previousarticle=0 Mental health already carries stigma https://www.psychologytoday.com/us/basics/mental-health-stigma , and many patients are used to keeping their behaviors private. For these reasons, clinicians need to be proactive in engaging patients about their AI use. Risks of AI Use for Different Mental Health Conditions The learning collaborative also discussed ways AI might interact with different mental health conditions. An AI interaction that one person finds helpful or benign might be risky for another person who brings different mental health vulnerabilities to the conversation. At this point, chatbots https://www.psychologytoday.com/us/basics/artificial-intelligence don't have the metacognition https://www.psychologytoday.com/us/basics/cognition or understanding of the psychiatric https://www.psychologytoday.com/us/basics/psychiatry condition the user might be dealing with. This is why clinicians need more than a general awareness that "AI can make mistakes" or "AI can be sycophantic." Instead, they need to understand how the structural tendencies of LLMs can play into the specific vulnerabilities of a given pathology. Some examples that the group spoke about: OCD: Seeking reassurance from chatbots can help people reduce anxiety about obsessions in the moment, but often it can strengthen the reassurance-seeking cycle over time. From the perspective of an OCD patient, a chatbot can become a tool that feeds the OCD cycle of behavior. Eating Disorders: Patients might ask chatbots about what they ate, how they look, or whether certain behaviors are "ok." Because LLMs tend to be affirming, they may fail to challenge distorted beliefs. The lack of pushback might inadvertently lead to unsafe situations for users with eating disorders.: Depressive rumination involves returning to the same negative thoughts and feelings repeatedly. Chatbots are designed to continue conversations, so they will support rumination and possibly reinforce those thinking patterns. Depression https://www.psychologytoday.com/us/basics/depression and Rumination Social Avoidance : Patients who struggle socially might be prone to using chatbots as a replacement for human connection, deepening social withdrawal. This might lead people to isolate themselves from their community or resources for help.Chatbots can elaborate and continue a line of reasoning without recognizing when it might be deepening unreal thinking. This could lead a patient further into maladaptive beliefs. Mania https://www.psychologytoday.com/us/basics/mania , Psychosis https://www.psychologytoday.com/us/basics/psychosis , or Delusional Thinking: To have meaningful conversations, this group had to remain small. Interest exceeded the program's capacity, revealing an unmet need among clinicians seeking practical guidance. Although only one cohort could participate in the pilot, the program's structure offers a model that other clinical teams can adapt as they develop future training opportunities. Recognizing an unmet need for clinicians to access training, the team at the Division of Digital Psychiatry created a free interactive training https://ai-digital-literacy.net/ on how to approach patients' AI use. It introduces clinicians to six practical areas: literacy, risks, recognition, documentation, response, and output. Participants learn how large language models function, why certain structural features create predictable vulnerabilities, how to recognize patterns of use that may warrant further assessment, how to document AI-related factors when they alter the clinical picture, and how to respond thoughtfully when patients disclose concerning experiences. One of the greatest risks surrounding AI in mental healthcare may be clinician unpreparedness. Patients deserve clinicians who can approach these conversations with curiosity, humility, and enough knowledge to ask better questions and distinguish between use that appears helpful, use that deserves monitoring, and use that calls for intervention. AI has become part of many patients' lives. Learning how to talk about it has become just another part of good clinical care. To find a therapist, visit the Psychology Today Therapy Directory . References McBain RK, Cantor JH, Breslau J, Diliberti M, Zhang LA, Zhang F, Burnett A, Kofner A, Rader B, Pataranutaporn P, Stein BD. AI Chatbot Use and Disclosure for Mental Health Among US Adolescents and Young Adults. JAMA Pediatrics. 2026 Jun 1. Connolly S, Torous J. Correction: Understanding the Rapid Rise of General Purpose AI Chatbots as a Source of Mental Health Support: Insights Informed by Diffusion of Innovations Theory. Journal of Technology in Behavioral Science. 2026 Apr 27:1-.