Abstract
<title>Abstract</title> <p>Background Generative artificial intelligence (GenAI) is entering health professions education faster than curricula and governance can adapt. Educational adoption depends not only on technical capability but also on whether students and educators consider generated explanations, feedback and learning support useful, trustworthy and ethically acceptable. Methods A rapid systematic review was conducted and reported using PRISMA 2020 and rapid-review guidance. Structured searches of PubMed-indexed records, ERIC, Scopus-indexed publisher platforms and backward/forward citations covered 30 November 2022 to 29 July 2026. English-language, peer-reviewed empirical studies were eligible when they examined health professions students’ or educators’ perceptions, attitudes, trust, acceptance, concerns or educational use of GenAI. One reviewer screened, extracted and appraised all records, followed by a complete verification pass against source records. Methodological limitations were assessed with design-specific Mixed Methods Appraisal Tool criteria. Because measures and populations were heterogeneous, findings were synthesised narratively and thematically without meta-analysis. Results Sixty-five records were identified, 14 duplicates removed, 51 unique records screened and 30 full texts assessed. Twenty studies involving approximately 9,553 participants were included: 4 published in 2023, 7 in 2024, 8 in 2025 and 1 in 2026. Most were cross-sectional surveys and 16 focused predominantly on students. Acceptance was positive but conditional. Learners valued rapid explanations, summarisation, brainstorming, feedback, accessibility and productivity; educators recognised similar potential. Trust remained task-dependent and weakened when outputs affected clinical reasoning, grading or professional accountability. Recurrent concerns were hallucination and misinformation, plagiarism, over-reliance, erosion of critical thinking, privacy, bias and unclear institutional rules. Prior experience and AI literacy generally predicted more positive perceptions, whereas educators and clinicians tended to be more cautious than students. Common methodological limitations were convenience sampling, self-report measures, limited nonresponse analysis and rapidly ageing model-specific findings. Conclusions Health professions stakeholders do not simply accept or reject GenAI; they calibrate acceptance according to educational stakes, expected accuracy and human oversight. GenAI is best positioned as a supervised learning aid rather than an autonomous authority. Curricula should combine AI literacy, verification skills, explicit assessment rules and educator accountability. More longitudinal, experimental and low- and middle-income-country research is required.</p>