JMIR Publications

ABSTRACT

Background:

Artificial intelligence (AI) is increasingly being integrated into healthcare workflows, with growing interest in AI-enabled clinical decision support systems (CDSSs). In mental health care, these tools may support screening, triage, monitoring, and treatment planning, but adoption is shaped by clinician trust, perceived usefulness, ethical considerations, and concerns about safety and confidentiality.

Objective:

This scoping review aims to synthesise current evidence on mental health clinicians’ attitudes, perceptions, and beliefs regarding the use of AI tools within mental health care, including perceived benefits, risks, and prerequisites for adoption.

Methods:

A scoping review was conducted in accordance with JBI guidance and PRISMA-ScR. Searches of six databases (CINAHL, Embase, PsycINFO, PubMed, Scopus, Web of Science) were completed on 20 October 2025 for studies published from 2020 onward. Eligible studies included clinicians in psychiatry or mental health settings and reported measures of attitudes, perceptions, or beliefs about AI (including AI-enabled CDSSs).

Results:

Twenty-three studies were included, with most published from 2024 onward. Overall, clinicians demonstrated cautious optimism toward AI, particularly when positioned as augmenting rather than replacing clinical expertise. Perceived benefits centred on reducing administrative burden (e.g., drafting/summarising documentation) and supporting synthesis of large volumes of patient data for screening, triage, and monitoring. However, clinicians commonly reported limited AI literacy and minimal real-world use beyond low-risk tasks. Key barriers included concerns regarding privacy, governance, data ownership, clinical safety and reliability (including risk of inaccurate or insensitive outputs), impacts on therapeutic relationships, overreliance, and uncertainty regarding medico-legal accountability. Across studies, clinicians emphasised prerequisites for adoption including education and training, clear guidelines, strong governance, role clarity, and ongoing human oversight.

Conclusions:

Mental health clinicians are broadly receptive to the potential of AI-enabled CDSSs, but current adoption is constrained by limited familiarity and substantial concerns about ethics, privacy, safety, accountability, and preservation of the therapeutic relationship. Implementation efforts should prioritise governance frameworks, clinician co-design, and staged introduction beginning with low-risk applications, alongside training that supports safe interpretation and use. Future research should also examine consumer perspectives to inform acceptable and effective integration of AI into mental health decision support.