Updated on Mar. 23, 2026 | 7 minute read
When chatbots enter therapy, how should we talk about them?
Katherine Gibb, M.Ed.
Filtered by tag
Perspectives
2 perspectives
Updated on Nov. 19, 2025 | 10 minute read
General versus Purpose-Built Models
Hassan Rezaee, Ph.D., Katherine Gibb, M.Ed.
Research
Showing 4 summaries.
Despite strong performance on medical licensing exams, LLMs may still fall short in real-world settings, where accuracy depends on interaction quality, such as both user input and the model's ability to communicate effectively to the user.
Andrew M. Bean, Rebecca Elizabeth Payne, Guy Parsons, et al.
Nature Medicine
Reviewing psychotherapy standards and testing both general LLMs and commercial therapy bots, the authors argue that current AI systems should not replace therapists because they can express stigma, respond inappropriately in some mental health situations, and cannot reproduce key human elements of care.
Jared Moore, Declan Grabb, William Agnew, et al.
Preprint
MindEval shows state-of-the-art models struggle with realistic multi-turn therapy conversations.
Jose Pombal, Maya D'Eon, Nuno M. Guerreiro, et al.
Preprint
A proposed framework suggests that general-purpose AI chatbots could perpetuate OCD and anxiety by reinforcing avoidance and reassurance-seeking.
Ashleigh Golden, Elias Aboujaoude
npj Digital Medicine
Editor's note: Includes a practical section on addressing reassurance-seeking dynamics in recurrent conversations with LLM chatbots.