Focal, intervention-specific, and titrated: Which artificial intelligence tools are FIT for mental health care?
TL;DR
FIT is a proposed framework, designed to reduce three specific risks that have been linked to conversational MH-AI, and the dimensions are intended as practical starting guardrails, not as an exhaustive list.
Abstract
For artificial intelligence (AI) used in conjunction with psychotherapy, clinicians and clients need a principled basis for deciding which tools to engage with, and intervention designers need guidance on building tools that are most likely to be safe and effective. As the evidence base on mental health (MH)-AI safety and efficacy grows, we propose tools that are FIT: focal (bounded in problem focus), intervention-specific (bounded in therapeutic procedures), and titrated (with dose and duration calibrated to match the use case). FIT is a proposed framework, designed to reduce three specific risks that have been linked to conversational MH-AI. The dimensions are intended as practical starting guardrails, not as an exhaustive list. In this viewpoint article, the authors' argument concerns only MH-AI used in routine clinical care-tools prescribed by or used with input for guidance from a clinician. The authors leave aside fully unsupervised consumer use, which raises additional concerns beyond this viewpoint. (PsycInfo Database Record (c) 2026 APA, all rights reserved).