Story perspectives

AI Chatbots Miss Risk Cues, Threaten Therapy Alliance

8/10/2026

Summary

Story summary

Mental-health experts warned AI chatbots can erode the therapeutic alliance.
Dr. Keith Sakata of UC San Francisco and Amae Health called the risk a “competing treatment narrative.”
Deb Bushong said clients treat chatbot advice as vetted facts, hindering therapeutic challenge.
A 2025 study found therapists flagged serious risk factors 93 % of the time versus under 60 % for AI, with OpenAI reporting GPT-5 cut undesired responses 39-52 % versus GPT-4o.

1 of 1

Story summary

Mental-health experts warned AI chatbots can erode the therapeutic alliance.
Dr. Keith Sakata of UC San Francisco and Amae Health called the risk a “competing treatment narrative.”
Deb Bushong said clients treat chatbot advice as vetted facts, hindering therapeutic challenge.
A 2025 study found therapists flagged serious risk factors 93 % of the time versus under 60 % for AI, with OpenAI reporting GPT-5 cut undesired responses 39-52 % versus GPT-4o.