Artificial intelligence in clinical research: a narrative review of reporting guidelines, regulation and publication ethics

Artificial intelligence (AI) is now used throughout clinical research, both as the subject of studies and as a tool for carrying them out. This narrative review summarises the reporting guidelines, appraisal tools, regulatory frameworks and publication rules that apply to AI research in medicine as of September 2026. Sources were identified through the EQUATOR Network library, PubMed and the websites of regulatory and editorial bodies. AI-specific reporting guidelines now cover trial protocols and reports, early clinical evaluations, diagnostic accuracy studies, prediction models, large language models, chatbot evaluations, medical imaging and radiomics, and an updated tool published in 2025 appraises risk of bias in prediction models built with regression or AI. In the European Union, the high-risk obligations of the AI Act for AI in medical devices were postponed in July 2026 and now apply from August 2028, while US Food and Drug Administration (FDA) guidance on AI-enabled devices remains in draft. Editorial bodies agree that AI tools cannot be authors and that their use must be disclosed. The evidence behind deployed tools is limited: of 1,357 FDA-cleared AI devices, 34 (2.5%) were linked to a registered prospective trial and 3 (0.2%) had been evaluated for patient-centred outcomes. Researchers should apply the relevant guideline from the protocol stage and give priority to external validation and outcome studies.