Can a chatbot stand in for a therapist?

No, a chatbot should not stand in for a therapist, and European law has no harmonized rule for who answers when its advice goes wrong.

This is for people who are between therapy sessions or still waiting for a first appointment, and for clinicians whose clients now arrive with advice from an AI tool. I checked the sources below in October 2026, starting with the World Psychiatry article. Together they show how chatbots performed in tests, how two papers frame responsibility, and how often people use chatbots for emotional topics. They do not show that any chatbot product causes harm in general, and they do not say how a court would rule.

The next useful step: if you need a therapist, start at couchcommons.com/find. If you are a licensed clinician, you can list at couchcommons.com/join.

This is general information, not therapy or medical advice.

Key takeaways

• Researchers who study this say chatbot conversations usually carry lower stakes than human relationships. World Psychiatry
• In tests, general-purpose and therapy-specific chatbots underperformed human therapists in chat scenarios involving mental health conditions. World Psychiatry
• Regulatory compliance does not decide whether psychiatric care was reasonable, according to a European viewpoint. JMIR
• Clinicians should remain responsible for contextual, patient-centered judgment. JMIR

Can a chatbot replace a therapist?

No. The evidence points one way. The World Psychiatry article reports that chatbots underperformed human therapists in chat scenarios involving mental health conditions. It also says human relationships involve stakes, mutual accountability, and social capital, and that chatbot interactions usually carry lower stakes. Read the article

If you act on chatbot advice, the sources do not say who answers for the result. That gap is the reason to bring any AI suggestion to a person who can answer for it. The World Psychiatry article puts the difference in terms of stakes and mutual accountability, and the JMIR viewpoint says clinicians should keep responsibility for contextual, patient-centered judgment. That is not a reason to avoid every AI tool. It is a reason to treat chatbot output as one input, not as the final word.

Who is responsible when chatbot advice goes wrong?

The evidence does not settle this. The JMIR viewpoint says European instruments "do not provide a harmonized fault-based liability framework for AI-assisted health care." It proposes that accountability be distributed across developers, deployers, and clinicians according to control and preventability. The authors offer that model as normative proposals, not settled European legal standards. Read the viewpoint

Are people already using chatbots for emotional support?

Yes. Anthropic reported in June 2025 that only 2.9% of Claude.ai interactions were affective conversations. That is a small share, and the post says it reflects adult usage patterns. A 2026 Nature Mental Health paper says millions now use these systems for emotional support and companionship. It also says individuals with preexisting mental health conditions may face increased risks of chatbot-induced changes in beliefs and behaviour.

What does a licensed therapist provide that a chatbot does not?

The sources point to two things: judgment and accountability. The viewpoint places responsibility for contextual, patient-centered judgment with clinicians. The World Psychiatry article names mutual accountability as part of what human relationships involve. When you look for a clinician, ask what they are licensed to do and how they handle a question that a chatbot would answer in seconds. You can start at couchcommons.com/find.

Takeaway

A chatbot can answer questions, but the evidence does not show that it can stand in for a therapist. Responsibility when its advice goes wrong is still being worked out, so treat chatbot advice as information, not as care.

Frequently asked questions

Can a chatbot replace a therapist?

No. In tests cited by World Psychiatry, chatbots underperformed human therapists in chat scenarios involving mental health conditions. The same article says human relationships involve stakes and mutual accountability, and that chatbot interactions usually carry lower stakes.

Who is responsible when chatbot advice goes wrong?

The evidence does not settle it. A viewpoint in the Journal of Medical Internet Research says European instruments do not provide a harmonized fault-based liability framework for AI-assisted health care. It suggests sharing responsibility among developers, deployers, and clinicians, and the authors call that a proposal.

Are people already using chatbots for emotional support?

Yes. Anthropic reported in June 2025 that 2.9% of Claude.ai interactions were affective conversations. A 2026 Nature Mental Health paper says millions of people now use these systems for emotional support and companionship.

What should I do if I am in crisis?

If you are in crisis, call or text 988 (US). A chatbot is not a substitute for that call.