AI‐Enhanced Point‐of‐Care Diagnostics for Infectious Diseases in Resource‐Limited Settings: A Scoping Review

Objectives To systematically map the extent and nature of research on AI‐enhanced point‐of‐care (POC) and rapid diagnostic technologies for infectious diseases in resource‐limited settings, and to identify gaps in disease coverage, geographic representation and validation rigour. Methods This scoping review followed JBI methodology and PRISMA‐ScR guidelines. The protocol was registered on OSF ( https://doi.org/10.17605/OSF.IO/KV8MP ). Five databases (PubMed, Embase, Scopus, Web of Science and IEEE Xplore) were searched for studies published from January 2015 to March 2026. Title/abstract and full‐text screening used rule‐based keyword screening with manual validation (Cohen's kappa = 0.856). Data were extracted using a 19‐variable charting form and enriched with PubMed Central full texts. Results From 1072 records, 551 remained after deduplication and 237 studies were included. Publication volume grew exponentially, with 44% published in 2025–2026. COVID‐19 (32%), malaria (27%) and tuberculosis (14%) dominated; neglected tropical diseases accounted for fewer than 8%. Microscopy (21%), molecular diagnostics (17%), biosensors (14%) and rapid diagnostic tests (14%) were the most common modalities. Convolutional neural networks predominated (26%), followed by random forests (10%) and support vector machines (8%). Only 7% of studies reported prospective field validation, while 62% did not report validation level. Geographic analysis revealed concentration in East Africa and South Asia, with underrepresentation of West Africa and Latin America. Conclusions AI‐enhanced POC diagnostics for infectious diseases in resource‐limited settings is a rapidly growing field facing critical gaps in validation rigour, disease equity and geographic representation. Only 16 of 237 studies (6.8%) report prospective field validation. Future research should prioritise field validation, expand beyond the COVID‐19/malaria/TB triad and involve end‐user communities from the design stage.