Artificial Intelligence in Healthcare: 2025 Year in Review

BackgroundArtificial intelligence research in healthcare continues to expand rapidly, with 2025 marking a transition from a text-centric Large Language Model and traditional Deep Learning model focus toward multimodal foundation models. To support the community in tracking these advancements, we conducted a comprehensive qualitative and quantitative ‘Artificial Intelligence Year in Review 2025.’ This dataset, utilized in the Year In Review, is released to provide a standardized, annotated repository of the high-maturity research that defined the field over the past year.MethodsOn January 1, 2026, PubMed database was queried for the terms "machine learning" or "artificial intelligence" and "2025," with the search restricted to English-language human-subject research. A BERT-based deep learning classifier, pre-trained and validated on manually labeled data, assessed publication maturity from initial results. Additionally, systematic reviews, duplicates, pre-prints, robotic surgery publications, and non-human research were excluded. Five reviewers then manually annotated publications for healthcare specialty, data type, and model type. Publications employing foundation models, including LLMs, Large Vision Models (LVM), and Multimodal Models (FM-MM), were further utilized and included in the dataset.ResultsOur PubMed search yielded 49,394 total articles of which 3,366 were classified as mature, and 2,966 publications were used for analysis and are made available after final exclusions. The dataset reveals a slight decline in classical Machine Learning (173) and a surge in Multimodal Foundation Models (144) when compared to the previous year in review. Imaging remains the dominant specialty (976), though the dataset captures a diversification into Administrative (277) and General (251) categories. Data type distribution shows Image data at 53.9% and Text at 38.2%, with a newly significant emergence of Audio data (1.2%). This dataset tracks the specific trajectories of various specialties, such as Oncology and Surgery, as they adopt higher-capacity foundation models.ConclusionThe 2025 dataset captures an inflection point in healthcare AI, documenting the doubling of research volume and the pivotal shift toward multimodal AI models. It is intended for researchers, policy-makers, and clinicians to identify trends and gaps in the current state of healthcare AI research and publications.Data Files DescriptionOne data file is provided, which illustrates the annotations of the mature sources used in the 2025 AI in healthcare review publication. The file includes 'DOI', 'Title', 'Abstract', 'Speciality', 'Model', 'Data type', 'PMID', 'Authors', 'Journal', 'First author', 'Citation', 'Year', 'Create Date'. The ‘Speciality’, ‘Model’, and ‘Data Type’ were manually annotated by the BrainXAI research team.