Why AI Chatbots Should Not Replace a Therapist

An AI chatbot can sound supportive without being qualified to provide mental health care. A “therapist” name, reassuring language, or an instruction to act like a psychologist does not establish clinical training, a professional license, or evidence that the tool is safe and effective.

The concern is especially important when people use general-purpose or character-based chatbots instead of seeking professional assessment. Conversation quality and treatment quality are different things.

What researchers found

A 2025 study involving Stanford, Carnegie Mellon, the University of Minnesota Twin Cities, and the University of Texas at Austin evaluated language models against expectations for therapists. The researchers’ report described inappropriate crisis responses, stigmatizing responses to some mental health conditions, and failures to challenge delusional beliefs appropriately.

These findings concern the models and scenarios tested. They do not establish that every chatbot behaves identically or that every digital mental health intervention is ineffective. They do show why fluent conversation is insufficient evidence that a chatbot can replace a clinician.

Why a convincing response can be unsafe

• Missed context: a chatbot may overlook warning signs, misunderstand a situation, or give advice based on incomplete information.
• Excessive agreement: validation can become harmful if it reinforces an inaccurate or dangerous belief instead of questioning it carefully.
• Unverified expertise: a fictional character’s claim to be licensed is not a credential you can independently verify.
• Limited accountability: a conversation with software does not create the same professional relationship, assessment process, or continuing responsibility as clinical care.

The American Psychological Association’s health advisory urges users to understand these differences and warns that evidence does not support treating general-purpose generative chatbots as qualified mental health providers. TipsMake’s coverage of research on AI counseling risks discusses related concerns.

Privacy and dependence deserve attention

Do not assume a chatbot conversation has the confidentiality protections of a clinical consultation. Check the service’s rules for storage, human review, training, deletion, and sharing before entering intimate information. The chatbot data-settings guide explains one part of that review; opting out of training is not the same as eliminating all data handling.

Notice whether chatbot use is displacing real relationships or professional care. Feeling understood by an app does not demonstrate that your symptoms are being assessed or that your situation is improving.

Keep any use supplementary

If you already use a chatbot, discuss that use with a qualified mental health professional. A limited task such as organizing questions for an appointment is different from asking a bot to diagnose a condition, select treatment, or make medication decisions.

If you or someone else is in immediate danger, contact local emergency services or an appropriate crisis service rather than waiting for a chatbot response. A wellness app or simulated therapist should not be your emergency plan.