Patient-facing artificial intelligence in primary health care: a scoping review of literature through early 2025

Le résumé fourni par la source

Abstract Background Patient-facing artificial intelligence (AI) tools are increasingly accessible to the public, but evidence on their development and impact in primary health care (PHC) remains limited. This study aimed to map the literature on patient-facing AI tools in PHC, including study characteristics, tool types, patient-journey stage, AI lifecycle stage, public availability, and health-system outcomes. Methods This scoping review was conducted in accordance with the PRISMA Extension for Scoping Reviews (PRISMA-ScR) reporting guideline. The search strategy was developed with a medical librarian and preregistered on the Open Science Framework. MEDLINE, Embase, CINAHL, Cochrane Library, CENTRAL, Web of Science, Google Scholar, ClinicalTrials.gov, and medRxiv were searched through March 5, 2025. Eligible studies were original research on patient-facing AI tools in PHC reporting on health-system or technical outcomes. Data were extracted and synthesized descriptively. Results Of 3,469 records identified, 65 studies published between 2016 and 2025 met inclusion criteria. The most common study designs were validation ( n = 20; 31%) and usability or user-centred design studies ( n = 13; 20%). The most common tool types were chatbots ( n = 31; 48%), mHealth apps ( n = 24; 37%), and symptom checker or triage tools ( n = 21; 32%), most often targeting pre-visit triage ( n = 43; 66%) and self-management ( n = 42; 65%). Twenty-seven tools (42%) were publicly accessible at the time of review. Reported outcomes focused on people-centredness ( n = 41; 63%) and technical performance ( n = 29; 45%), with limited evaluation of patient safety ( n = 5; 8%) and access ( n = 4; 6%). Conclusions The literature on patient-facing AI tools in PHC has been concentrated on early-stage evaluations of patient experience and technical performance, with limited evaluation of safety and access. Given that a substantial proportion of these tools are already publicly accessible, robust pre- and post-deployment evidence is urgently needed, supported by institutions that uphold rigorous evaluation across the AI lifecycle.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.