A persistent mismatch between rising mental health demand and a shrinking clinical workforce has made digital augmentation not just convenient but structurally necessary. This scoping review offers one of the most comprehensive mappings yet of where AI-driven robots are actually being deployed in psychiatric and psychological care — and what technologies are powering them.
Drawing on seven major databases and spanning January 2017 through May 2025, the review synthesized evidence from 34 studies. AI robots were mapped across five functional domains: diagnosis (19 studies), treatment and intervention (27 studies), counseling (13 studies), risk prediction and identification (8 studies), and monitoring (5 studies). The conditions most represented were anxiety disorders, depressive disorders, autism spectrum disorder, and dementia. Machine learning algorithms emerged as the core computational engine in diagnostic applications, though the review's narrative synthesis approach — rather than meta-analysis — means effect-size comparisons across studies were not attempted. The JBI scoping methodology with PRISMA-ScR reporting adds methodological transparency, though 34 included studies across such heterogeneous conditions and robot types limits how definitive conclusions can be.
This review arrives at a pivotal moment: social robots, conversational AI agents, and embodied virtual assistants are converging into a single product category, yet the evidence base remains fragmented across engineering, clinical psychology, and psychiatry literatures. The heavy representation of treatment and counseling functions — 40 combined studies — suggests the field is moving beyond passive monitoring toward active therapeutic roles, a shift with significant implications for clinical governance and liability. The autism and dementia focus reflects populations where human-robot interaction may reduce stigma and improve engagement, though generalizability to broader adult mental health populations is unproven. For longevity-focused readers, the dementia applications are particularly relevant, as early-stage cognitive monitoring via AI companions could extend independent living. The field remains largely at proof-of-concept stage; most included studies were small, heterogeneous in design, and short in duration — meaning clinical adoption should proceed with measured expectations.
