Artificial Intelligence for Continuing Medical Education: Applications, Challenges and Prospects

Continuing Medical Education (CME) is defined as structured post‑licensure continuing professional education targeting practising, licensed clinicians and frontline healthcare practitioners, whose core purpose is to maintain and upgrade clinical competence for real‑world patient care. For this review, CME is strictly confined to post‑licensure professional education for actively practising clinicians and healthcare providers; undergraduate medical education and pre‑registration training for medical students fall outside the scope of this analysis. Continuing Medical Education (CME) enables clinicians and healthcare staff to update professional knowledge, master emerging technologies and adapt to evolving clinical practice. Amid rapid medical innovation and frequent updates of clinical practice guidelines, frontline practitioners increasingly demand flexible, targeted and efficient training resources. Nevertheless, traditional CME often adopts unified teaching modes, consumes substantial time and fails to satisfy differentiated clinical requirements [1]. Diverse AI technologies, encompassing machine learning, natural language processing and intelligent simulation systems, have created new opportunities to reshape CME via personalized learning, automated evaluation and evidence‑based training design [2]. This article conducts a narrative review of practical applications of AI in CME, analyses prominent obstacles, and prospects developmental trends, aiming to deliver insights for innovation in AI‑powered continuing professional education for practising health workers.

Personalized adaptive learning

AI platforms integrate learners’ historical CME learning records, professional titles, departmental practice backgrounds, and measured clinical competency gaps to generate tailored learning trajectories. Targeted courses, real‑world clinical cases and up‑to‑date clinical guidelines are intelligently recommended to improve learning efficiency and practical relevance, guiding practicing clinicians to priorities high‑priority knowledge and skills [3]. Current CME‑specific evidence supports the technical feasibility of adaptive recommendation pipelines; however, adequately powered interventional trials demonstrating tangible improvements in learning outcomes among licensed clinicians remain scarce. Large language models further support on‑demand knowledge consultation and self‑directed assessment, accommodating fragmented learning scenarios for busy frontline practitioners [4].

Despite these practical advantages, generative large language models pose distinct safety risks for CME workflows, including factual hallucinations, obsolete guideline content, prompt‑dependent output instability and automation bias. Given the lack of sufficient empirical data specifically validated in formal CME practitioner cohorts, the safety characteristics and potential risks of large language models summarized in this section are partially supported by general medical education research findings. Since AI‑generated information may directly influence clinical decision‑making in patient care, mandatory human expert validation must be embedded for all outputs used within CME programmes.

AI‑enhanced clinical simulation training

AI‑driven virtual‑patient platforms, intelligent image analysis workflows, and robot‑mediated simulation deliver repeatable, immersive training without exposing real patients to clinical risk [5]. These tools can reinforce clinical reasoning, hand‑on operational proficiency and crisis‑management performance among practicing clinicians, with particular utility in neurology, critical‑care practice and diagnostic imaging interpretation [6]. Available CME‑focused observational work demonstrates improvements in simulation‑based performance; robust evidence confirming skill transfer to real‑world patient‑care settings has not yet been established. Due to the limited number of large-scale empirical studies focusing specifically on formal CME scenarios, the conclusion that generative artificial intelligence can dynamically adjust simulated clinical scenarios according to clinician competency and CME teaching objectives is partially derived from general health education research, which provides a theoretical and practical reference for CME application [7].

Intelligent assessment and competency evaluation

AI algorithms automatically process CME examination responses, analyse clinical case deliberations and quantify the quality of clinical decision‑making among participating practitioners. This technology delivers near‑instant feedback and flags potential competency deficits to support standardized CME evaluation [8]. Claims of fully objective, bias‑free AI‑driven assessment require important caveats within CME settings. Although AI may reduce certain forms of human rater subjectivity, new algorithmic biases can originate from imbalanced training datasets, model design flaws, or uneven performance across clinician subgroups. At present, systematic reliability and validity benchmarks for AI based clinical competency assessment are mainly verified in general medical education research and targeted validation studies for formal licensed clinician CME systems are still lacking. When compared against manual evaluation workflows, AI‑supported assessment can partially mitigate subjective scoring variance and reduce administrative workload for CME educators [9].

Data‑driven training program design

By mining multisource real‑world datasets including clinical workload metrics and local disease‑spectrum statistics, AI‑based analytical workflows identify high‑priority training requirements for clinician cohorts and optimize CME curriculum modules. Most existing literature describes analytical frameworks; large‑scale real‑world deployment studies embedded within official CME systems remain limited. This analytical approach narrows the disconnect between CME curricula and frontline clinical realities and strengthens the practical value of continuing‑education activities. CME administrative bodies can leverage these analytical outputs to avoid redundant or misaligned training provisions for practicing healthcare professionals.

The widespread adoption of AI in CME faces multiple substantial hurdles. First, aggregation of CME learning logs and de‑identified clinical datasets raises notable data‑privacy and cybersecurity risks [10]. Second, limited model interpretability erodes clinical educators’ confidence in algorithm‑derived outputs used for practicing clinicians’ competency appraisal [11]. Third, unified, field‑specific evaluation benchmarks for AI‑enabled CME platforms have not been established [12]. Fourth, algorithmic bias may produce inequitable learning recommendations for different subgroups of CME participants [13]. This risk includes factual artefacts produced by generative‑AI tools and performance heterogeneity across diverse clinician populations, as elaborated in Subsection 2.1. In addition, interoperability gaps between newly‑developed AI systems and legacy hospital and CME information infrastructures impede large‑scale real‑world implementation [14].

Future advancement of AI‑augmented CME should prioritise multicentre prospective validation of AI platforms intended for practising clinicians, promote research on explainable artificial intelligence tailored to continuing‑education contexts, and develop unified evaluation criteria for CME‑oriented AI tools. These criteria ought to cover technical robustness, CME‑specific educational validity, algorithmic‑bias mitigation, and explicit safety safeguards such as mandatory human‑in‑the‑loop expert review for generative‑AI outputs. Integration of 5G connectivity, virtual‑reality modules and robotic simulation may enable hybrid intelligent CME training ecosystems. Policymakers, clinical departments and CME‑delivering organisations must jointly formulate context‑specific ethical frameworks governing AI deployment within formal continuing‑medical‑education programmes.

A critical, under‑investigated research gap lies in the causal pathway linking AI‑supported CME participation to downstream clinical practice modification and patient‑centred outcomes. Most current studies only measure proximal learning metrics such as quiz scores, completion rates or simulation‑task performance; these surrogate endpoints cannot confirm whether knowledge and skills acquired via AI‑CME are transferred into daily clinical workflows. Future longitudinal multicentre research should adopt mixed‑method study designs that sequentially evaluate three linked outcome tiers: (1) immediate CME‑level learning gains (knowledge acquisition, self‑reported competency improvement); (2) intermediate clinical practice outcomes, including adherence to clinical guidelines, diagnostic accuracy, appropriate therapeutic decision‑making and documentation quality extracted from electronic health records; (3) distal patient‑level endpoints such as complication rates, readmission rates, treatment response and patient‑reported experience metrics [6].

Such tiered evaluation frameworks help disentangle whether observed learning improvements remain confined to the educational environment or translate into tangible real‑world clinical benefits. Additionally, implementation research is needed to identify organizational, individual, and technical barriers that interrupt this knowledge‑to‑practice transition, including clinician workload burden, digital literacy gaps, lack of institutional incentives for applying newly‑acquired competencies, and trust‑related barriers toward AI‑generated educational content. Establishing validated linkage mechanisms between CME platform datasets and de‑identified hospital clinical registries will be essential to support this line of outcome‑oriented CME research.

Artificial intelligence offers potential to improve personalisation, efficiency and clinical relevance for continuing medical education among licensed practising clinicians, yet high‑quality real‑world empirical evidence for many AI‑CME use‑cases remains sparse. Through innovation across personalised learning, clinical simulation, competency evaluation and data‑informed curriculum development, AI creates new avenues to sustain lifelong professional learning for medical practitioners. Addressing outstanding challenges in data governance, domain‑specific ethical standards and cross‑system interoperability will further unlock AI’s practical value for CME, supporting healthcare‑system improvement and sustained enhancement of clinical service quality.

1. Charow R, Jeyakumar T, Younus S, Dolatabadi E, Salhia M, et al. (2021) Artificial intelligence education programs for health care professionals: Scoping review. JMIR Med Educ 7(4): e31043.
2. Hlede V, Valanci S, D'Antuono GR, Dow H, O'Beirne R, et al. (2026) AI‑enhanced continuing professional development as an evolving sociotechnical system: Multimethod theoretical framework development study. JMIR Med Educ 12(2): e69156.
3. Hughes S (2026) Building AI capability in medical imaging: A codesigned continuing professional development framework and evaluation. BMC Med Educ 26(1): 192.
4. Lyle CR (2026) Leveraging large language models to advance certification, physician learning, and diagnostic excellence. J Contin Educ Health Prof 46(1): 34‑
5. Ye MZ, Kwok TT, Wong JYH, Jiang J (2026) GenAI‑supported virtual patients in health care education: Systematic review. J Med Internet Res 28(5): e82756.
6. Al‑Qaaneh AM, Qunaibi EA, Ismail BF, Muhidat HI, Rageh FS, et al. (2026) Effectiveness of informed AI use on clinical competence of general practitioners and internists: Pre‑post intervention study. JMIR Med Educ 12(1): e75534.
7. Yamamoto A, Koda M, Ogawa H, Miyoshi T, Maeda Y, et al. (2024) Enhancing medical interview skills through AI‑simulated patient interactions: Nonrandomized controlled trial. JMIR Med Educ 10(1): e58753.
8. Fazal A, Carter M (2025) Artificial intelligence in surgical education: A 2025 update on adaptive training, feedback, and competency‑based education. Am Surg 91(11): 1397-
9. Liu Y, Chen T (2026) A survey on medical competence evaluation benchmarks for large language models. Health Inf Sci Syst 14(1): 50.
10. Li X, Yan X, Lai H (2025) The ethical challenges in the integration of artificial intelligence and large language models in medical education: A scoping review. PLoS One 20(10): e0333411.
11. Hou Z, Chen J, Guo H (2026) From “teaching by word and deed” to “intelligent mentorship”: Ethical reconsiderations of AI‑enabled medical education. Front Med 12: 1754139.
12. Park S, Kim J (2025) Technologies, opportunities, challenges, and future directions for integrating generative artificial intelligence into medical education: A narrative review. J Educ Inform 9(2): 77‑
13. Fabiyi O, Zehra F, Tariq M, Abidi SH, Almazan J (2026) AI‑supported case‑based learning in medical education: A comprehensive scoping review. Front Med 13: 1798097.
14. Richards N (2026) Ethical considerations for teaching with artificial intelligence: A scoping review in medical education settings. Teach Learn Med 38(2): 189-198.