When Therapy Clients Use AI Chatbots

Clients are often reluctant to bring up chatbot use, expecting disapproval, so the clinician usually has to ask first. Useful questions are concrete: which tool, how often, at what times of day, what the client asks it, and how they feel afterward. Asking to see a sample conversation, with the client's permission, often reveals more than a summary does. There are real benefits. A chatbot is available at 2 a.m., costs little and feels low-stakes. Clients use it to rehearse assertive requests, draft a difficult message, get psychoeducation in plain language, or organize journaling. Used this way it can extend the work done in session. The risks follow from how these systems behave. General chatbots tend to agree with and validate the user, which can harden a distorted belief, whether that is certainty about a partner's betrayal or a conviction that nobody cares. They can answer the same anxious question endlessly, which suits compulsive reassurance-seeking perfectly. Heavy companion-app use can replace human contact instead of supporting it. Conversations are not covered by therapist confidentiality and may be stored or reviewed by the company. Crisis handling is inconsistent, and families have filed lawsuits alleging chatbots contributed to harm to vulnerable users. Some US states have passed laws limiting AI from providing therapy. A common misconception is that the clinician's job is to forbid chatbot use. A harm-reduction stance usually works better: agree on helpful uses, name the unhelpful patterns, and treat the chatbot as one more behavior in the case formulation. Another misconception is that a client who uses a chatbot values therapy less; often the chatbot fills gaps between sessions that the client could not otherwise fill. Clinicians should also avoid entering client information into chatbots themselves.