As digital tools become more prevalent, mental health providers are increasingly encountering patients who utilize artificial intelligence (AI) platforms, such as general chatbots or wellness applications, between their scheduled therapy sessions.
Because client-initiated use of these technologies can impact the therapeutic process, this article offers an objective framework to help therapists navigate these disclosures, assess clinical risks, and maintain standard boundaries of care.
Understanding the landscape of AI use
It’s helpful to look at recent data regarding when, why, and how individuals are independently engaging with AI to understand the implications for clinical dynamics.
- Patient-reported engagement: A randomized clinical trial published in JAMA Network Open (1) evaluated trends among individuals experiencing psychological distress who chose to use conversational AI agents. The data show that some users reported a reduction in self-reported anxiety and depression symptoms, noting a perceived digital therapeutic alliance with the software.
- Night-time use: A massive, de-identified analysis of over 500,000 health conversations published in Nature Health (2) demonstrated that consumer use of AI chatbots for emotional well-being and symptom assessment peaks drastically during evening and nighttime hours, indicating users are turning to AI when human support systems and traditional healthcare hours are most limited.
- Unexpressed clinical narrative: A clinical commentary in JAMA Psychiatry (3) highlights that many patients intentionally use AI to confess potentially stigmatizing thoughts, intrusive symptoms, or fears they feel too embarrassed or ashamed to initially share with a human clinician. Asking about a client's AI usage can provide a rich window into their unexpressed clinical narrative.
- Clinical distinction: While users report high satisfaction with AI responsiveness, the American Psychological Association (APA) (4) emphasizes a critical distinction: generative AI lacks genuine clinical reasoning, diagnostic capabilities, and emotional reciprocity.
Whether we view these tools with skepticism or curiosity, recent data suggest that client-initiated AI use is an active clinical variable. It can shape a patient’s self-understanding, diagnostic assumptions, and symptom presentation long before their initial appointment with a therapist.
A patient-centered framework for exploring AI use
Integrating questions about AI into practice does not have to be an endorsement; but rather a tool for deeper clinical understanding. Adapted from the patient-centered screening guidelines outlined in JAMA Psychiatry (3), therapists can utilize a conversational framework to collaboratively explore a client's AI use.
Step 1: Normalize and ask
Focus on reducing stigma to unlock unexpressed clinical narratives that the client may feel too embarrassed or stigmatized to share face-to-face.
“I’ve noticed that many people lately are turning to AI platforms like ChatGPT to process stress or look for coping strategies between sessions. I'm curious if you've ever experimented with any digital tools like that for support or guidance?”
Step 2: Explore motivation and perceived benefits
Identifying the client's underlying therapeutic needs by noting timelines or motivation for engaging with AI can specifically pinpoint gaps in the client's distress tolerance skills or uncover key clinical insights such as sleep hygiene skills or areas of concern.
“What specific needs are you usually hoping to meet when you engage with AI, and what aspects of the responses have felt the most supportive to you?”
Step 3: Explore client concerns and ambivalence
Evaluate the client’s digital insight and assess for experiential avoidance, noticing if the client is using the safe, risk-free feedback of AI to retreat from the vulnerability required in human relationships.
“On the flip side, have you noticed any instances where AI’s feedback felt off, unhelpful, or perhaps left you feeling more anxious or misunderstood?”
Step 4: Offer information with permission
With the client's permission, we can step in as educators to intercept algorithmic misinformation and bridge clinical knowledge gaps. This is our opportunity to address several key AI limitations outlined by the APA (4):
- Over-validation: AI is engineered to please the user and maximize agreement. It rarely challenges cognitive distortions or highlights maladaptive patterns, which can inadvertently create an echo chamber that reinforces symptoms.
- Inaccurate outputs: AI operates on statistical pattern recognition, not truth, and is highly prone to "hallucinations" (generating false information with absolute confidence).
- Confidentiality and safety: Standard platforms lack HIPAA protections, meaning personal data may be stored by commercial entities. It’s important to reiterate that AI chatbots are not a suitable replacement for crisis intervention or safety planning.
Step 5: Invite ongoing dialogue
Conclude by establishing that the client's digital interactions are welcome in the therapeutic space, signaling that this is an ongoing clinical dialogue rather than a one-time screening, and recenters the human alliance.
“I’d love for us to treat this as an ongoing conversation rather than a one-time check-in. Please feel entirely free to bring in any interesting prompts or responses you receive between sessions so we can look at them together.”
Core principles
Rula respects the clinical decision-making of our therapist network. Discussing a client’s use of AI does not require adopting or approving of the technology. Rather, applying clinical curiosity, boundary-setting, and risk assessment that therapists have always used to protect patient well-being in an ever-changing world can lead to improved dialogue with clients and help guide the therapeutic interaction safely forward.
References
- JAMA Network Open: Evaluation of Conversational AI Agents and Perceived Alliance in Psychological Distress.
https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2847751 - Costa-Gomes, B., et al. (2026). Public use of a generalist LLM chatbot for health queries. Nature Health.
https://www.nature.com/articles/s44360-026-00117-x - Saba, S. K., & Weeks, W. B. (2026). Patients Use AI—Clinicians Should Ask How. JAMA Psychiatry.
https://jamanetwork.com/journals/jamapsychiatry/fullarticle/2847068 - APA Resource Hub (2026). Guide to Navigating AI in Behavioral Health.
https://www.apa.org/topics/artificial-intelligence-machine-learning/guide-navigating-ai
Updated