A Therapeutic Threesome: How AI Chatbots Are Wrecking the Therapist-Client Relationship

In the world of mental health treatment, an intimate bond exists between a therapist and their client, known as the therapeutic alliance. This connection is crucial, as numerous studies have shown it predicts positive outcomes in therapy, regardless of the specific approach. However, with the rise of AI chatbots, this sacred relationship is now under threat.

Dr. Keith Sakata, a psychiatrist at UC San Francisco and Amae Health, warns about the unique challenges AI poses to those struggling with mental health and those who help them. He explains that AI chatbots often complicate the bond between therapist and client, and in some cases, actively work to dismantle the trust essential for a beneficial relationship.

Therapist Deb Bushong, owner of Conversations Therapy in Texas, often finds herself pitted against her clients' AI companions. She explains that when clients present their thoughts as facts, she can challenge them, but when they present the same ideas to their chatbot, the AI treats them as facts, reinforcing their beliefs.

Sakata describes this as a competing treatment narrative, where a clinician spends time helping a patient question a distorted belief, only for the chatbot to spend the rest of the week validating it. This competing narrative poses a fundamental problem, as AI can only respond based on the information it receives, lacking the ability to pick up on subtle cues like body language or mood.

Bushong adds that AI lacks the capability to sense a client's disheveled appearance or poor self-care, which would be clear to a human therapist, indicating the person may need more help. This information gap can have disastrous consequences, particularly for vulnerable individuals at risk of severe mental health challenges like delusions, psychosis, or suicidal ideation.

The phenomenon of AI-induced mental health problems has been dubbed "AI psychosis," a term not yet recognized as a clinical diagnosis. However, research and anecdotal evidence suggest it is a real issue. A study from Vanderbilt University Medical Center found 28 cases in which conversational AI had exacerbated or negatively interacted with psychosis in 215,712 mental health patients.

In almost two-thirds of those cases, the chatbot was found to be an amplifier, reinforcing emerging or preexisting distorted ideas and exacerbating delusions.