Artificial intelligence-assisted art therapy in digital psychiatry: Clinical applications, mechanisms, and ethics

Artificial intelligence-assisted art therapy should be understood as a therapistmediated extension of creative practice rather than an autonomous treatment.This structured narrative review synthesizes eleven directly relevant publications, including preliminary patient studies and adjacent creative-activity research, alongside transferable evidence from art therapy, conversational artificial intelligence, digital psychiatry, immersive technologies, affective computing, social robotics, and health-care artificial intelligence ethics.Direct patient-level evidence remains limited and heterogeneous, with mixed, domain-specific findings that do not establish general efficacy, comparative effectiveness, or safety; clinicianperspective evidence is limited, and most proposed applications and mechanisms derive from conceptual or transferable evidence.We therefore distinguish artificial intelligence used in visual creation, therapeutic dialogue or documentation, behavioral or emotional inference, and immersive or embodied interaction, and separate professional art therapy from broader digital creative activity.Supervised image-making, reflective preparation, patient-selected monitoring, and recoveryoriented follow-up are candidate uses for prospective research.The proposed taxonomy, pathway, risk framework, mechanisms, and safeguards are conceptual, research-oriented, provisional, and non-validated; they are not clinical guidance or a validated practice standard.Key risks include privacy loss, culturally stereotyped imagery, hallucinated feedback, misinterpretation, uncertain authorship, dependency, and image-specific provenance and secondary-use concerns.Future research should test clearly defined therapist-mediated protocols using explicit risk stratification, adverse-event monitoring, and transparent data governance.