When Clients Bring ChatGPT Diagnoses to Therapy: What the Data Shows and How to Respond

By The Professions Cloud Editors Updated November 2026 · 5 sources · 5 min read

Prices and facts checked November 2, 2026. We don't accept payment for rankings.

The trend: clients increasingly arrive at therapy having already talked to an AI chatbot about their problems, and sometimes with a diagnosis it suggested:

• In the American Psychological Association's 2026 Chatbots and Mental Health Survey of 1,242 US licensed psychologists, 77% said patients had discussed using AI for support or other mental health purposes. 39% had patients who used AI to self-diagnose (APA via Medical Xpress).
• At a counselling conference on October 29, 2026, a psychotherapist reported that 65% of his practice's new clients seeking help for emotional or addiction concerns in 2026 had already used AI chatbots (GlobeNewswire via The Manila Times). That's one practice's own records, not a study, but it matches what many clinicians describe.

Why it matters: if you don't ask about AI use, you're missing part of the picture: what your client has been told, what they've disclosed and where they turn between sessions. This guide covers what the data says and how to handle it in session.

What clients are using chatbots for

From the APA survey, fielded April 9–26, 2026 (APA via Medical Xpress, Decrypt):

The good and the bad. Among psychologists whose patients had ongoing chatbot conversations, 68% said patients felt supported or validated, and 41% said patients used chatbots to reinforce healthy coping skills. But 36% saw patients develop dependency on a chatbot, and 15% saw distorted thinking or delusions linked to chatbot use.

What psychologists worry about:

• 97% say chatbots may reinforce negative behaviors or delusional beliefs.
• 94% say today's chatbots can't treat conditions with appropriate nuance.
• 89% worry they may inadvertently encourage self-harm.
• Still, 54% are comfortable with some patients using them, and 40% are optimistic chatbots could help when a professional isn't available.

The APA notes that the survey only captures what psychologists' existing patients told them, so it likely understates how widely people use chatbots for mental health.

What the APA says about AI and therapy

The APA's position is that AI isn't a safe or effective replacement for a qualified mental health provider (Decrypt). Its health advisory separates higher-risk and lower-risk uses (Stacker):

• Higher risk: diagnosing conditions, deciding whether medication is needed, managing suicidal thoughts and guiding someone through a crisis.
• Lower risk: guided breathing, general stress information and sorting through thoughts before a session.

The APA's advice to the public includes checking any AI mental health information with a practitioner, asking for research-backed strategies, prompting the AI to challenge their thinking, and limiting use so it doesn't crowd out sleep, work and relationships (APA via Medical Xpress).

How to talk about it in session

1. Ask every client, without judgment. Add a question to intake and check-ins, such as "Do you use ChatGPT or any other AI apps to talk through how you're feeling?" Many clients won't mention it unless asked.
2. Treat an AI diagnosis as a starting point. A suggested label like "anxiety" or "bipolar" often comes from a single conversation, without the history-taking a real assessment needs (The Manila Times). Ask what led them to it and what it meant to them, then do your own assessment.
3. Ask to see the conversation. If the client is willing, reading what the chatbot said shows you the advice they're following and what they've shared. It may also reveal risk they haven't raised with you.
4. Screen for dependency and isolation. Ask how often they use it, whether it's replacing people, and whether they tell the AI things they tell no one else.
5. Check for crisis use. Ask whether they've turned to a chatbot when feeling unsafe. Make sure their safety plan includes human contacts and crisis lines, not just an app.
6. Talk about privacy. Consumer chatbots aren't covered by HIPAA. Clients may not realize what they've shared and how it can be stored or used.
7. Make it useful where you can. For some clients, a chatbot can help with low-risk tasks between sessions, such as practicing a skill, journaling prompts or preparing for a hard conversation. Agree on boundaries together.
8. Document it. Note significant AI use, any AI-suggested diagnoses the client holds, and any risk concerns, as you would other relevant history.

Where the law stands

Several states now restrict companies from offering AI as therapy, and some require chatbots to send users in crisis to human help. That governs the companies, not your clients' personal use. See our 2026 map of state AI chatbot laws and AI companion chatbot laws. For what California did, see California's SB 903 veto.

The scale is large. OpenAI has said about 1.2 million ChatGPT users a week share suicidal thoughts with it (CalMatters). For the longer view on your profession, see will AI replace therapists?

FAQ

How many therapy clients use AI chatbots?

In the APA's 2026 survey of 1,242 psychologists, 77% said patients had discussed using AI for mental health purposes, and 39% had patients who used AI to self-diagnose. The APA says this likely understates use, because it only counts what patients told their psychologists.

What should I do if a client has an AI-generated diagnosis?

Treat it as useful information, not a conclusion. Ask how they reached it and what it means to them, then complete your own assessment. If they're willing, read the chatbot conversation together.

Does the APA say clients shouldn't use AI chatbots?

The APA says AI isn't a safe or effective replacement for a qualified provider. Its advisory flags diagnosis, medication decisions and crisis support as higher-risk uses, while breathing exercises and sorting through thoughts before a session are lower risk.

Should I ask clients about AI use at intake?

Yes. Many clients won't mention it unless asked, and their chatbot use can affect what they believe about their condition, what they've disclosed and how they cope between sessions.

Changelog: Nov 2, 2026: first published.