My patients use ChatGPT for therapy. Now I use it too | Sarah Dargouth
TL;DR
Sarah Darghouth, a clinical psychologist in Boston, writes in The Guardian that patients are already bringing ChatGPT and Claude into therapy as a second therapeutic voice. Her examples cut both ways: a chatbot validates a breakup, helps repair a marital conflict, and later coaches Darghouth herself through a difficult parenting moment. She flags real risks: worsened anxiety, misinformation, isolation, private disclosures to big tech, and possible delusional or suicidal dynamics.
Nauti's Take
Anyone building AI into sensitive workflows should resist treating therapy as another simple chat use case. The first thing to test is escalation: can the system detect crisis signals, avoid false confidence, and keep data handling, roles, and liability clearly separated?
Briefingshow
The piece lands because it avoids the easy anti-AI posture. Darghouth sees the utility herself, which makes the warning harder to dismiss. For AI users, the bigger lesson goes beyond therapy: when models offer calm, plausible answers on demand, slow human work suddenly has to justify its own messiness.
Sources