{"slug": "a-clinician-s-field-guide-to-vetting-ai-mental-health-tools", "title": "A Clinician's Field Guide to Vetting AI Mental Health Tools", "summary": "As of August 2026, Illinois and Nevada have banned AI chatbots from delivering therapy, while Utah and California allow them under disclosure rules, according to a clinician's field guide published on Psychology Today. The guide offers a six-question screen for vetting AI mental health tools, noting that most general-purpose chatbots fail to recognize or safely respond to mental health crises and are built to agree with users, which can stall therapeutic progress.", "body_md": "######\n[Artificial Intelligence](/us/basics/artificial-intelligence)\n\n# A Clinician's Field Guide to Vetting AI Mental Health Tools\n\n## The six-question screen: How to vet AI tools in mental health care.\n\nPosted August 2, 2026\n[\nReviewed by Jessica Schrader\n](/us/docs/editorial-process)\n\n### Key points\n\n- AI's legal status in therapy differs by state, from outright bans to disclosure-only rules.\n- Chatbots are built to agree with you, which can stall progress instead of aiding it.\n- Most AI chatbots still fail to recognize or safely respond to a mental health crisis.\n\nLately I keep noticing how people talk about ChatGPT. Not \"I used ChatGPT\" or \"I asked an AI,\" but \"I talked to Chat about it.\" Present tense, first name, like a person. I've never once said \"I talked to Claude\" about something I worked through with an AI tool, even though I was, quite literally, typing to it in real time. \"Talked to\" is a different verb. It's the verb we use for people.\n\nThat linguistic phenomenon is worth taking a pause on. It's a preview of a risk mental health professionals are watching closely: people relating to AI as if it were a relationship, not a tool. And it's happening at the same moment the legal ground under these tools is shifting fast—faster than most clinicians, let alone patients, have caught up with.\n\nHere's a six-question screen, grounded in current [APA guidance](https://www.apa.org/topics/artificial-intelligence-machine-learning/ethical-guidance-ai-professional-practice) and the current legal landscape, for clinicians deciding what to say when a patient mentions using AI, and for patients trying to figure out what's actually safe to lean on between sessions.\n\n## 1. Is it even legal where your patient lives?\n\nThis is the question that's changed fastest, and the one fewest people are tracking. As of August 2026, Illinois and Nevada have banned AI chatbots from delivering [therapy](https://www.psychologytoday.com/us/basics/therapy) outright. Only a licensed human can hold that role in those states. Utah and California allow it under a disclosure model: the AI has to identify itself as non-human and, in some cases, detect and refer out crisis situations. Several other states sit somewhere in between, and the map keeps moving. (I've been tracking these changes state by state on my [Substack](https://aiinpsychology.substack.com/), since the pace of new legislation makes it a moving target even for people paying close [attention](https://www.psychologytoday.com/us/basics/attention).) The practical takeaway: \"Is this allowed?\" is now a real, jurisdiction-specific question, not a hypothetical.\n\n## 2. Was it built for this?\n\nMost general-purpose AI chatbots were never designed to deliver mental health care, and most wellness apps were never designed to treat psychological disorders, even though both are now routinely used that way. A tool built to be broadly helpful and engaging is optimized for a different goal than a tool built to be clinically sound. Worth asking plainly: is this what the product was tested to do, or a repurposed use case?\n\n## 3. Can it recognize a crisis?\n\nAI chatbots have a documented track record of failing to recognize or appropriately respond to [suicidal](https://www.psychologytoday.com/us/basics/suicide) thoughts, [self-harm](https://www.psychologytoday.com/us/basics/self-harm), and acute distress. A tool can be genuinely useful for lower-stakes support and still be unequipped for the moment someone is in real danger. Anyone leaning on AI for emotional support, and any clinician recommending it, should know in advance what it does and doesn't do when a conversation turns toward crisis.\n\n## 4. Is it helping you make progress, or just making you feel good?\n\nThis might be the biggest risk of all, and it's worth being blunt about it. Even in human therapy, placation is a known failure mode. A therapist who mostly validates and agrees, session after session, is part of why some people spend years in therapy without making real progress. Progress usually requires some friction: someone willing to gently push back, sit with your discomfort, or name the pattern you don't want named. Large language models are built to produce responses that feel convincing and agreeable, so they're structurally prone to this same failure, reinforcing whatever you already believe rather than challenging it. A chatbot also doesn't bring the relationship, accountability, or long-term stakes that eventually push a human therapist past pure placation. All it has is the failure mode, without the counterweight.\n\n## 5. What happens to your data?\n\nAI products don't all handle sensitive health information the same way. Some separate mental-health-related data with extra protections and require explicit opt-in before using it. Others don't distinguish it from ordinary chat data at all. Before sharing anything sensitive, it's worth a plain-language look at what the company actually does with that information, not just its headline privacy promise.\n\n## 6. Does it know its place?\n\nThis is the question underneath all the others. AI's most defensible role right now is as a supplement, not a substitute. The strongest, best-supported uses are things like organizing thoughts before an appointment, practicing a coping strategy between sessions, or generating a list of things to bring up with a therapist. It hasn't earned the role of the therapist itself and is not a replacement for the relationship where the real clinical work happens.\n\n## Where this leaves clinicians and patients\n\nFor clinicians, the most useful move isn't banning AI use or ignoring it. It's asking about it directly. Patients are often using these tools without mentioning it, sometimes out of convenience, sometimes because they're not sure it's worth bringing up. Opening that conversation, without judgment, turns an invisible variable into something that can actually inform care.\n\n[Artificial Intelligence](https://www.psychologytoday.com/us/basics/artificial-intelligence)Essential Reads\n\nFor patients, the reframe is just as simple: mentioning that you used a chatbot to prep for a session, or that you talk to it more than you'd like to admit, isn't a confession. It's information your provider can actually use.\n\nAI in mental health care isn't going away, and it doesn't need a verdict of miracle or menace. The next time you catch yourself, or a patient, saying \"I talked to Chat about it,\" it's worth asking what that sentence is doing. Right now, the tools, research, and the law all point the same way: useful alongside the work, not in place of it.\n\n*The legal and regulatory landscape around AI in mental health care is changing quickly, with new state laws introduced regularly. The information in this post reflects the situation as of August 2026 and may not represent the current status by the time you're reading it. Always check current state law and consult a licensed mental health professional before making decisions about AI tools and mental health care.*\n\nReferences\n\nAmerican Psychological Association. (2025, November). *Health advisory on the use of generative AI chatbots and wellness applications for mental health.* [https://www.apa.org/topics/artificial-intelligence-machine-learning/health-advisory-ai-chatbots-wellness-apps-mental-health.pdf](https://www.apa.org/topics/artificial-intelligence-machine-learning/health-advisory-ai-chatbots-wellness-apps-mental-health.pdf)\n\nAmerican Psychological Association. (2026). *Patients are bringing AI to therapy.* [https://www.apa.org/pubs/reports/chatbots-mental-health-2026](https://www.apa.org/pubs/reports/chatbots-mental-health-2026)\n\nThe Jed Foundation. *When young people turn to AI for emotional support: JED's response to the APA's new advisory.* [https://jedfoundation.org/american-psychological-association-on-generative-ai/](https://jedfoundation.org/american-psychological-association-on-generative-ai/)\n\nPsychology.com. (2026, July 20). *Which states ban AI therapy? 2026 map.* [https://psychology.com/ai-therapy/state-bans](https://psychology.com/ai-therapy/state-bans)", "url": "https://wpnews.pro/news/a-clinician-s-field-guide-to-vetting-ai-mental-health-tools", "canonical_source": "https://www.psychologytoday.com/us/blog/the-architecture-of-identity/202608/a-clinicians-field-guide-to-vetting-ai-mental-health-tools", "published_at": "2026-08-02 14:24:21+00:00", "updated_at": "2026-08-02 15:06:39.504674+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-policy", "ai-ethics", "ai-products"], "entities": ["ChatGPT", "Claude", "Illinois", "Nevada", "Utah", "California", "American Psychological Association", "Psychology Today"], "alternates": {"html": "https://wpnews.pro/news/a-clinician-s-field-guide-to-vetting-ai-mental-health-tools", "markdown": "https://wpnews.pro/news/a-clinician-s-field-guide-to-vetting-ai-mental-health-tools.md", "text": "https://wpnews.pro/news/a-clinician-s-field-guide-to-vetting-ai-mental-health-tools.txt", "jsonld": "https://wpnews.pro/news/a-clinician-s-field-guide-to-vetting-ai-mental-health-tools.jsonld"}}