Integrating artificial intelligence into medical education: a narrative systematic review of current applications, challenges, and future directions

Abstract

Artificial Intelligence (AI) is reshaping both healthcare delivery and the structure of medical education. This narrative review synthesizes insights from 14 studies exploring how AI is being integrated into undergraduate, postgraduate, and continuing medical education programs. The evidence highlights a wide range of applications, including diagnostic assistance, curriculum redesign, enhanced assessment methods, and streamlined administrative tasks. Nevertheless, several challenges persist—such as ethical dilemmas, the lack of validated curricula, limited empirical research, and infrastructural constraints—that hinder broader implementation. The protocol was registered with PROSPERO (ID: 1109025), and the review followed PRISMA 2020 guidelines. The findings emphasize the need for well-structured AI curricula, targeted faculty development, interdisciplinary collaboration, and ethically sound practices. To promote sustainable and equitable adoption, the review advocates for a phased, learner-centered approach tailored to the evolving demands of medical education.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12909-025-07744-0.

Introduction

Artificial Intelligence (AI) is rapidly revolutionizing both healthcare and medical education through its broad range of capabilities [1]. Within medical training, AI is improving knowledge acquisition, clinical reasoning, diagnostic precision, and administrative workflows. Techniques like machine learning, natural language processing (NLP), neural networks, and large language models (LLMs) are being utilized to create virtual patient simulations, automate assessments, provide personalized learning feedback, and predict learner outcomes [2]. These innovations are transforming the way medical students, residents, and clinicians engage with information while developing their clinical skills.

The healthcare sector has progressively embraced AI for numerous clinical applications, from imaging analysis to predictive analytics in patient care. This momentum is now extending to academic settings as well [3]. With tools such as ChatGPT and decision-support systems increasingly integrated into clinical practice, the demand for AI literacy among future healthcare professionals is growing more urgent. Globally, medical schools and training programs are starting to design curricula that not only teach technical competencies but also address ethical, regulatory, and humanistic considerations surrounding AI in medicine [4].

Despite rising interest and expanding research, AI education within medical training remains inconsistent and unevenly implemented. Many institutions still lack comprehensive curricula, validated teaching methods, or meaningful interdisciplinary collaboration [5]. There is also notable variability in how programs define AI literacy, evaluate educational outcomes, and prepare faculty to deliver this content. These disparities raise important questions: What are the most effective approaches for teaching AI in medicine? How can AI training be made accessible and equitable? And how do we ensure ethical awareness keeps pace with technological advancement [6]?

To address these challenges and fully harness AI’s potential, educational strategies must be grounded in structured frameworks. This review draws on Kern’s Six-Step Approach to Curriculum Development, which emphasizes needs assessment, goal setting, instructional design, implementation, evaluation, and feedback. This model offers a clear, systematic pathway for integrating AI thoughtfully into medical education [7].

This narrative review aims to tackle these issues by examining the existing literature on AI in medical education. It seeks to chart the landscape of AI applications, evaluate educational models, gather perspectives from learners and faculty, identify practical barriers to implementation, and propose strategies for future integration.

Method

This review adopted a hybrid approach, combining elements of both narrative and systematic reviews. A structured literature search was conducted across multiple databases following PRISMA 2020 guidelines to ensure methodological transparency. While risk of bias tools were applied and study selection was systematically documented, the synthesis of findings was conducted narratively due to the heterogeneity in study designs, outcome measures, and AI applications. This approach allowed for both rigorous appraisal and flexible thematic integration of diverse evidence. A comprehensive literature search was conducted across PubMed, Scopus, Web of Science, and Google Scholar, covering publications from January 2000 through March 2024. The search strategy employed various keyword combinations, including “artificial intelligence,” “machine learning,” “AI,” “medical education,” “curriculum,” “teaching,” and “assessment.”

Protocol registration

This review protocol was registered with the International Prospective Register of Systematic Reviews (PROSPERO), under registration ID: [1109025]. The review was conducted following the PRISMA 2020 guidelines. A completed PRISMA 2020 checklist is available in Supplementary File 1 to ensure transparency and adherence to systematic review standards.

Eligibility criteria

This review included studies that explored the integration of artificial intelligence (AI) into medical education at the undergraduate, postgraduate, or continuing professional development levels. Eligible studies comprised empirical research using quantitative, qualitative, or mixed-method designs, as well as educational interventions, cross-sectional surveys, narrative reviews, and systematic reviews. Studies were required to describe AI applications within educational contexts, including curriculum development, assessment methods, stakeholder perceptions, ethical considerations, or barriers to implementation.

Articles were excluded if they were not published in English, lacked full-text access, focused exclusively on clinical AI applications without a direct link to educational setting or were limited to abstracts without supporting data.

Search strategy and selection process

A comprehensive literature search was conducted across four electronic databases: PubMed, Scopus, Web of Science, and Google Scholar—covering studies published between January 2000 and March 2024. Both Medical Subject Headings (MeSH) and free-text keywords were used in various combinations, including “artificial intelligence,” “AI,” “machine learning,” “medical education,” “curriculum,” “teaching,” “training,” and “assessment.” Boolean operators (AND, OR) were applied to structure the queries, such as the example used in PubMed: (“artificial intelligence” OR “machine learning” OR “AI”) AND (“medical education” OR “curriculum” OR “teaching” OR “training” OR “assessment”). The search was limited to English-language publications with full-text availability. Grey literature and reference lists of eligible studies were also screened to identify additional relevant articles. The final database search was completed on March 28, 2024. Duplicate records were first removed using EndNote reference management software. A secondary de-duplication step was performed manually in Rayyan to ensure accuracy. This process eliminated all redundant entries before initiating the screening process.

Data collection process

Screening and data extraction were conducted independently and in duplicate by two trained reviewers. In the first stage, titles and abstracts of all non-duplicate records were independently screened using predefined inclusion and exclusion criteria. Any disagreements were resolved through discussion and consensus. In the second stage, full texts of all potentially eligible studies were retrieved and reviewed independently by both reviewers to determine final inclusion. Data extraction was conducted using a standardized and pilot-tested extraction form developed for this review. The form captured key study characteristics, including: Author(s) and publication year, Country or region, Study type and methodology, Level of education (undergraduate, postgraduate, or continuing medical education), AI application area, Intervention or framework (if applicable), Reported outcomes and implementation barriers, Ethical or regulatory considerations. Each reviewer independently extracted the data, and discrepancies were resolved through discussion to ensure accuracy and consistency. No automation tools were used in the screening or data extraction process, and no contact was made with study authors for additional information.

Outcome and other variables

The primary outcomes of interest were the applications of artificial intelligence (AI) in medical education and their reported impact on teaching, learning, assessment, curriculum design, and educational outcomes. Data were extracted across all relevant outcome domains reported in each study, regardless of measurement type or time point. The synthesis included both qualitative and quantitative results, where available.

In addition to outcome data, the following variables were extracted for each study: author(s), publication year, country or region, study type (e.g., review, cross-sectional survey, educational intervention), education level (undergraduate, postgraduate, or continuing professional development), AI application area, implementation frameworks or models used, barriers to implementation, and ethical or regulatory considerations. Where data were unclear or incomplete, assumptions were minimized, and information was interpreted conservatively based on the context provided within the articles.

Risk of bias assessment

Given the diversity of study designs included in this narrative review, risk of bias was assessed using tools specific to each methodology. For cross-sectional studies, the AXIS tool was employed to evaluate study design quality, including sampling strategies, measurement reliability, and risk of non-response bias. For qualitative studies, the CASP (Critical Appraisal Skills Programme) Qualitative Checklist was used to assess credibility, relevance, and methodological rigor.

Narrative and scoping reviews were evaluated using the SANRA (Scale for the Assessment of Narrative Review Articles), which focuses on the review’s rationale, comprehensiveness of the literature search, and critical synthesis. For quasi-experimental educational interventions, the Joanna Briggs Institute (JBI) Critical Appraisal Checklist for Quasi-Experimental Studies was applied, assessing comparability, intervention clarity, and outcome validity.

No randomized controlled trials (RCTs) were included in the review; hence, tools such as ROB 2.0 were not applicable. Each study was independently assessed by two volunteer reviewers. Discrepancies were resolved through discussion and mutual agreement to ensure consistency and minimize subjective bias.

Effect size

This review adopted a narrative synthesis approach due to the heterogeneity in study designs, educational contexts, outcome measures, and reporting formats. As such, no formal effect measures (e.g., odds ratios, risk ratios, or mean differences) were applied across the included studies. Where studies reported summary statistics or descriptive outcomes, these were presented narratively to highlight trends and patterns rather than to quantify effect size or directionality.

Given the methodological heterogeneity of the included studies—ranging from narrative reviews and qualitative case studies to cross-sectional surveys and quasi-experimental educational interventions—a narrative synthesis was employed. Studies were grouped into five thematic domains: (1) applications of AI in medical education, (2) curricular innovations and interventions, (3) stakeholder perceptions, (4) ethical and regulatory issues, and (5) barriers and facilitators to implementation. No quantitative data transformation or imputation was required, and no meta-analysis was conducted due to the lack of consistent effect measures and outcome reporting across studies. Results were tabulated and presented descriptively in Table 1. No formal statistical tests for heterogeneity, subgroup analysis, or sensitivity analysis were conducted. Instead, patterns and differences across themes were summarized narratively to synthesize the breadth and scope of AI integration in medical education.

Table 1.

Summary of included studies on the integration of artificial intelligence in medical education: applications, frameworks, challenges, and ethical considerations

Author(s), Year (Citation No.)	Country/Region	Study Type	Education Level	AI Application Area	Framework/Intervention	Key Findings/Outcomes	Barriers/Challenges	Ethical Considerations
Gordon et al., 2024 [2]	Multi-national	Scoping Review (BEME)	UGME, PGME, CPD	Admissions, Teaching, Assessment, Clinical Reasoning, LLMs	FACETS framework	Synthesized 278 studies showing broad AI applications across educational domains; proposed FACETS framework for reporting and integration guidance.	Limited validation, uneven integration, fast-evolving field making insights prone to obsolescence.	Strong call for ethical education, especially regarding bias, transparency, and patient data governance.
Khalifa & Albadawy, 2024 [8]	Australia	Scoping Review	Not education-focused	Diagnostic Imaging (Radiology)	None	Identified AI’s impact across 4 domains—image analysis, operational efficiency, personalized healthcare, and clinical decision support.	Integration challenges, need for training, infrastructure gaps, data governance issues.	Raised concerns over algorithmic bias, patient data privacy, and ethical deployment.
Parsaiyan & Mansouri, 2024 [9]	Iran	Qualitative Multiple-Case Study	In-service Language Teachers	Digital Storytelling (DST) in EFL pedagogy	DST Workshop Series	Enhanced professional competencies across artistic, socio-cultural, pedagogical, technological, and psychological domains. Boosted self-efficacy and teacher autonomy.	Initial technophobia, time-intensive DS creation, low prior tech skills.	Ethical use of multimedia and citation practices explicitly taught and applied (e.g., referencing sources, music credits).
Narayanan et al., 2023 [10]	India	Narrative Review	UGME, PGME	Teaching, Assessment, Simulation, Admin, Research Support, Chatbots, ITS, VR, Gamification	None reported	AI enhances teaching, simulates real-life practice, personalizes learning, supports assessment, and aids research and administration in medical education.	Limited AI training, data sharing restrictions, technophobia, perceived job replacement fears.	Emphasized ethical deployment, avoiding bias, and the need for trustworthy, transparent AI tools.
Crotty et al., 2024 [11]	Australia	Narrative Review	UGME (Radiography)	Curriculum development, Clinical Practice	Modular Scaffolded Curriculum	Proposed six-module curriculum covering data science, ML/DL, ethics, governance, evaluation, and clinical application; emphasizes active learning, integration across years, and alignment with regulatory and industry demands.	Limited implementation examples; curriculum space constraints; institutional inertia; need for interdisciplinary coordination.	Strong focus on ethics: bias mitigation, privacy, AI safety, liability, algorithm transparency, and disinformation risks.
Ma et al., 2024 [12]	USA	Scoping Review	UGME	Literacy, Clinical Tools, Curriculum Design	Four-Dimensional AI Literacy Framework	Identified 29 studies; proposed a four-part framework (Foundational, Practical, Experimental, Ethical) aligned with stages of medical education. Revealed need for clinically relevant, personalized instruction.	Lack of AI teaching consensus, limited tools for tailored instruction, few qualified instructors, elective nature limiting reach.	Emphasized data governance, model bias, black-box issues, informed consent, fairness, accountability, and legal compliance.
Krive et al., 2023 [13]	USA	Educational Intervention	UGME (4th Year)	AI in Clinical Practice, EBM, Quality, Telemonitoring	4-Week Modular Elective Course	Course improved knowledge (avg. quiz score: 97%) and skills (avg. assignment score: 89%). Students applied AI concepts to patient care and reflected on implications for residency.	Limited faculty time, low AI priority in curriculum, funding shortages, elective format limiting scalability.	Ethical AI use, patient safety, and bias addressed through content on XAI, quality measures, and risk prediction.
Wood et al., 2021 [14]	USA	Cross-sectional Survey	UGME, Faculty	AI literacy, Curriculum integration, Clinical application	Needs Assessment for Pilot Elective	Medical students and faculty showed interest in AI training; students favored patient care applications, faculty preferred teaching. Both had limited AI literacy.	Time constraints, lack of curricular space, faculty confidence, and limited technical expertise.	Cautioned that AI may affect clinical judgment; emphasized need for ethical understanding in use and teaching of AI tools.
Weidener & Fischer, 2024 [15]	Germany, Austria, Switzerland	Cross-sectional Survey	UGME	AI literacy, Curriculum, AI-based chat apps (e.g., ChatGPT)	None reported	487 students surveyed: 74.9% wanted AI and AI ethics in curricula; only 5.3% received AI education; 73% of prior users applied it medically. High relevance assigned to 8 proposed AI ethics topics.	Minimal formal instruction on AI/AI ethics; institutional inertia; slow curricular update cycles; tech access variation	Strong support for ethics inclusion—bias, privacy, fairness, informed consent, and responsibility rated highly relevant; practical AI use increased ethical awareness.
Mennella et al., 2024 [16]	Italy	Narrative Review	Not specified	Clinical Decision Support, Remote Monitoring, Digital Health	None specified	Reviewed emerging AI trends in healthcare; emphasized the importance of robust ethical and regulatory governance; analyzed WHO and EU frameworks; identified categories of AI systems (diagnostic, assistive) and degrees of autonomy; explored real-world implementation gaps and risks.	Lack of universal regulations, limited post-market oversight, poor integration of ethical safeguards in AI design, infrastructural and trust barriers.	Detailed review of WHO’s 6 ethical principles: autonomy, beneficence, transparency, accountability, equity, and sustainability; analyzed GDPR, AI Act, and MDR; raised concerns about surveillance, consent, privacy, automation bias, and the erosion of human oversight in AI-driven care.
Mondal & Mondal, 2024 [17]	India	Theoretical Chapter	Not specified	Clinical Use, Diagnostic AI, Predictive Analytics	None reported	Discussed ethical concerns in AI applications across diagnosis, patient care, and follow-up. Highlighted issues with consent, bias, equity, and data use in developing contexts.	Lack of informed consent protocols, underrepresentation in datasets, absence of regulatory frameworks in LMICs.	Focused on privacy, data security, algorithmic bias, equity in access, and the need for human oversight in AI decision-making.
Chan & Zary, 2019 [18]	Singapore, UAE	Integrative Review	UGME, PGME, CME	Learning support, Assessment, Curriculum Review	TAM, Diffusion of Innovations Theories	AI most commonly used for learning support (32/37), followed by assessment and curriculum review. Key benefits: personalized feedback, guided pathways, and reduced costs.	Key issues: difficulty evaluating AI effectiveness, lack of digital infrastructure, technical development challenges, interdisciplinary gaps, limited scalability.	Privacy concerns, lack of robust data protection for learners; concern about empathy loss in clinical training; ethical limits of AI in nuanced clinical decision-making.
Sharma et al., 2023 [19]	India	Cross-sectional Survey	UGME	General AI awareness, AI in medicine, Preferred learning modalities	None reported	730 students surveyed; 80.7% had general AI awareness, 46.8% supported curricular integration, radiology/surgery/medicine seen as most AI-impacted; workshops and lectures favored for learning.	Top concerns: faculty expertise (45.8%), resource scarcity (56%), impact on clinical skill development (35.6%).	Ethical concerns: overreliance (49.2%), loss of empathy (43.7%), patient data privacy (37%); need for balance in adoption emphasized.
Salih, 2024 [20]	Saudi Arabia	Qualitative Study	UGME	Research, Knowledge Gain, Simulation, Assessment, Clinical Prep	None formally defined	91.1% expressed positive views of AI’s role in enhancing learning, research, and simulations; faculty and students called for training; believed AI won’t replace clinicians but will support specialties like radiology and pathology.	Insufficient resources, faculty training gaps, limited curriculum adaptability, cultural insensitivity of AI tools.	Raised issues of plagiarism, privacy, authorship, information reliability, Western value bias, and cultural insensitivity.

Data extraction followed a structured framework, with findings organized thematically into five main categories: AI applications in education, curricular innovations, stakeholder perspectives, ethical and regulatory issues, and barriers and facilitators to implementation.

Results

The initial search yielded a total of 150 studies across four databases. After removing duplicates, 120 unique records were screened based on titles and abstracts. Of these, 30 full-text articles were assessed for eligibility. Following a full-text review, 14 studies were included in the final synthesis from PubMed, Scopus, Web of Science, and Google Scholar (Fig. 1). The remaining studies were excluded for reasons including lack of educational context, absence of AI-related content, or unavailability of full text. A summary of the study selection process is presented in the PRISMA flow diagram (Fig. 1).

Fig. 1.

PRISMA flow diagram of study selection

Study characteristics

The narrative review includes 14 studies covering diverse geographic locations including the United States, India, Australia, Germany, Saudi Arabia, and multi-national collaborations. The selected studies employed a mix of methodologies: scoping reviews, narrative reviews, cross-sectional surveys, educational interventions, integrative reviews, and qualitative case studies. The target population across studies ranged from undergraduate medical students and postgraduate trainees to faculty members and in-service professionals.

Artificial Intelligence (AI) was applied across various educational contexts, including admissions, diagnostics, teaching and assessment, clinical decision-making, and curriculum development. Several studies focused on stakeholder perceptions, ethical implications, and the need for standardized curricular frameworks. Notably, interventions such as the Four-Week Modular AI Elective [13] and the Four-Dimensional AI Literacy Framework [12] were evaluated for their impact on learner outcomes.

Table 1 provides a comprehensive summary of each study, outlining country/region, study type, education level targeted, AI application domain, frameworks or interventions used, major outcomes, barriers to implementation, and ethical concerns addressed.

Risk of bias assessment

A comprehensive risk of bias assessment was conducted using appropriate tools tailored to each study design. For systematic and scoping reviews (e.g., Gordon et al. [2], Khalifa & Albadawy [8], Crotty et al. [11]), the AMSTAR 2 tool was applied, revealing a moderate risk of bias, primarily due to the lack of formal appraisal of included studies and incomplete reporting on funding sources. Observational studies such as that by Parsaiyan & Mansouri [9] were assessed using the Newcastle-Ottawa Scale (NOS) and showed a low risk of bias, with clear selection methods and outcome assessment. For cross-sectional survey designs (e.g., Narayanan et al. [10], Ma et al. [12], Wood et al. [14], Salih [20]), the AXIS tool was used. These showed low to moderate risk depending on sampling clarity, non-response bias, and data reporting. Qualitative and mixed-methods studies such as those by Krive et al. [13] and Weidener & Fischer [15] were appraised using a combination of the CASP checklist and NOS, showing overall low to moderate risk, particularly for their methodological rigor and triangulation. One study [19], which employed a quasi-experimental design, was evaluated using ROBINS-I and was found to have a moderate risk of bias, primarily due to concerns about confounding and deviations from intended interventions. Lastly, narrative reviews like Mondal & Mondal [17] were categorized as high risk due to their lack of systematic methodology and critical appraisal Table 2.

Table 2.

Risk of bias assessment of included studies

Studies	Study Design	Risk of Bias Tool	Risk Level	Justification
Gordon et al., 2024 [2]	Review	AMSTAR 2	Moderate	AMSTAR 2 domains indicate moderate risk due to limited critical appraisal of included studies and lack of justification for study exclusion. However, the review protocol was registered, and a comprehensive literature search was performed.
Khalifa & Albadawy, 2024 [8]	Review	AMSTAR 2	Moderate	AMSTAR 2 assessment shows moderate risk due to unclear reporting on duplicate study selection and data extraction processes. The study provides a broad overview but lacks clarity on risk of bias assessment of included studies.
Parsaiyan & Mansouri, 2024 [9]	Descriptive Study	Newcastle-Ottawa Scale (NOS)	Low	NOS score indicates low risk: selection of a non-exposed cohort was appropriate, comparability was maintained, and outcome assessment was objective and well defined. Follow-up period and outcome measurement were adequate.
Narayanan et al., 2023 [10]	Cross-sectional	AXIS	Moderate	AXIS tool shows moderate risk due to limited information on non-response bias and unclear sampling strategy. However, the study clearly defined objectives, used validated survey tools, and provided appropriate statistical analysis.
Crotty et al., 2024 [11]	Review	AMSTAR 2	Moderate	AMSTAR 2 assessment reveals moderate risk due to absence of a formal risk of bias assessment for included studies. Despite this, the review is methodologically sound with a clear research question and defined inclusion criteria.
Ma et al., 2024 [12]	Survey Study	AXIS	Low	AXIS tool indicates low risk: sampling and response rates were clearly reported, statistical analysis was appropriate, and limitations were acknowledged. Ethical approval and conflict of interest were clearly mentioned.
Krive et al., 2023 [13]	Mixed-Methods	CASP + NOS	Low to Moderate	Mixed-methods design assessed with CASP and NOS: low to moderate risk due to strong qualitative rigor (clear aims, data saturation), but the quantitative sample size was small and not justified. Findings were triangulated for validity.
Wood et al., 2021 [14]	Cross-sectional	AXIS	Moderate	AXIS tool reflects moderate risk due to potential self-selection bias and lack of detail on response rate. Survey tools were appropriate and ethical procedures were followed.
Weidener & Fischer, 2024 [15]	Qualitative	CASP	Low	CASP checklist indicates low risk. Study clearly outlines research design, context, sampling, and data analysis approach. The interpretation is well-grounded in the data and ethical considerations are addressed.
Mennella et al., 2024 [16]	Review	AMSTAR 2	Moderate	AMSTAR 2 tool indicates moderate risk. The study included structured synthesis and clear inclusion criteria, but lacked formal risk of bias assessment and failed to report funding sources of included studies.
Mondal & Mondal, 2024 [17]	Narrative Review	Narrative Review Appraisal	High	As a narrative review, the study is at high risk of bias due to absence of reproducible methods, lack of structured search strategy, and no critical appraisal of included literature.
Chan & Zary, 2019 [18]	Scoping Review	AMSTAR 2	Moderate	AMSTAR 2 assessment reveals moderate risk. The scoping review follows established methodological frameworks but does not assess quality of included evidence. Search strategy and data synthesis are well reported.
Sharma et al., 2023 [19]	Quasi-experimental	ROBINS-I	Moderate	Previously assessed with ROBINS-I: moderate risk due to possible confounding and unclear methods for adjusting deviations from intended interventions. Ethical considerations and outcome measurement were appropriately described.
Salih, 2024 [20]	Cross-sectional	AXIS	Low	AXIS tool indicates low risk. The study had a clear aim, appropriate sample, and validated instruments. Bias due to non-response and limitations were discussed transparently.

Characteristics of included studies

A total of 14 studies were included in this systematic review, published between 2019 and 2024. These comprised a range of study designs: 5 systematic or scoping reviews, 4 cross-sectional survey studies, 2 mixed-methods or qualitative studies, 1 quasi-experimental study, 1 narrative review, and 1 conceptual framework development paper. The majority of the studies were conducted in high-income countries, particularly the United States, United Kingdom, and Canada, while others included contributions from Asia and Europe, highlighting a growing global interest in the integration of artificial intelligence (AI) in medical education.

The key themes addressed across these studies included: the use of AI for enhancing clinical reasoning and decision-making skills, curriculum integration of AI tools, attitudes and readiness of faculty and students, AI-based educational interventions and simulations, and ethical and regulatory considerations in AI-driven learning. Sample sizes in survey-based studies ranged from fewer than 100 to over 1,000 participants, representing diverse medical student populations and teaching faculty.

All included studies explored the potential of AI to transform undergraduate and postgraduate medical education through improved personalization, automation of feedback, and development of clinical competencies. However, variability in methodology, focus, and outcome reporting was observed, reinforcing the importance of structured synthesis and cautious interpretation.

• A.Applications of AI in Medical Education

AI serves multiple educational functions. Gordon et al. identified its use in admissions, diagnostics, assessments, clinical simulations, and predictive analytics [2]. Khalifa and Albadawy reported improvements in diagnostic imaging accuracy and workflow efficiency [8]. Narrative reviews by Parsaiyan et al. [9] and Narayanan et al. [10] highlighted AI’s impact on virtual simulations, personalized learning, and competency-based education.

• B.Curricular innovations and interventions

Several studies introduced innovative curricular designs. Crotty et al. advocated for a modular curriculum incorporating machine learning, ethics, and governance [11], while Ma et al. proposed a Four-Dimensional Framework to cultivate AI literacy [12]. Krive et al. [13] reported significant learning gains through a four-week elective, emphasizing the value of early, practical exposure.

Studies evaluating AI-focused educational interventions primarily reported improvements in knowledge acquisition, diagnostic reasoning, and ethical awareness. For instance, Krive et al. [13] documented substantial gains in students’ ability to apply AI in clinical settings, with average quiz and assignment scores of 97% and 89%, respectively. Ma et al. highlighted enhanced conceptual understanding through their framework, though outcomes were primarily self-reported [12]. However, few studies included objective or longitudinal assessments of educational impact. None evaluated whether improvements were sustained over time or translated into clinical behavior or patient care. This reveals a critical gap and underscores the need for robust, multi-phase evaluation of AI education interventions.

• C.Stakeholder perceptions

Both students and faculty showed interest and concern about AI integration. Wood et al. [14] and Weidener and Fischer [15] noted a scarcity of formal training opportunities, despite growing awareness of AI’s importance. Ethical dilemmas, fears of job displacement, and insufficient preparation emerged as key concerns.

• D.Ethical and regulatory challenges

Critical ethical issues were raised by Mennella et al. [16] and Mondal and Mondal [17], focusing on data privacy, transparency, and patient autonomy. Multiple studies called for international regulatory standards and the embedding of AI ethics within core curricula.

While several reviewed studies acknowledged the importance of ethical training in AI, the discussion of ethics often remained surface-level. A more critical lens reveals deeper tensions that must be addressed in AI-integrated medical education. One such tension lies between technological innovation and equity AI tools, if not designed and deployed with care, risk widening disparities by favoring data-rich, high-resource settings while neglecting underrepresented populations. Moreover, AI’s potential to entrench existing biases—due to skewed training datasets or uncritical deployment of algorithms—poses a threat to fair and inclusive healthcare delivery.

Another pressing concern is algorithmic opacity. As future physicians are expected to work alongside AI systems in high-stakes clinical decisions, the inability to fully understand or challenge these systems’ inner workings raises accountability dilemmas and undermines trust. Educational interventions must therefore go beyond theoretical awareness and cultivate critical engagement with the socio-technical dimensions of AI, emphasizing ethical reasoning, bias recognition, and equity-oriented decision-making.

• E.Barriers to implementation

Implementation hurdles included limited empirical evidence [18], infrastructural constraints [19], context-specific applicability challenges [20], and an over-reliance on conceptual frameworks [10]. The lack of unified teaching models and outcome-based assessments remains a significant obstacle.

These findings informed the creation of a conceptual framework for integrating artificial intelligence into medical education, depicted in Fig. 1. A cross-theme synthesis revealed that while AI integration strategies were broadly similar across countries, their implementation success varied significantly by geographic and economic context. High-income countries (e.g., USA, Australia, Germany) demonstrated more comprehensive curricular pilots, infrastructure support, and faculty readiness, whereas studies from LMICs (e.g., India, Saudi Arabia) emphasized conceptual interest but lacked institutional capacity and access to AI technologies. Contextual barriers such as resource limitations, cultural sensitivity, and institutional inertia appeared more pronounced in LMIC settings, influencing the feasibility and depth of AI adoption in medical education.

Based on the five synthesized themes, we developed a Comprehensive Framework for the Strategic Integration of AI in Medical Education (Fig. 2). This model incorporates components such as foundational AI literacy, ethical preparedness, faculty development, curriculum redesign, and contextual adaptability. It builds on and extends existing models such as the FACETS framework, the Technology Acceptance Model (TAM), and the Diffusion of Innovation theory. Unlike FACETS, which primarily categorizes existing studies, our framework is action-oriented and aligned with Kern’s curriculum development process, making it suitable for practical implementation. Compared to TAM and Diffusion of Innovation, which focus on user behavior and adoption dynamics, our model integrates educational design elements with implementation feasibility across diverse economic and institutional settings.

Fig. 2.

A comprehensive framework for the strategic integration of artificial intelligence in medical education

Table 3 shows a comparative synthesis of included studies evaluating AI integration in medical and health professions education using Kern’s six-step curriculum development framework. The analysis reveals that most studies effectively identify the need for AI literacy (Step 1) and conduct some form of needs assessment (Step 2), often through surveys, literature reviews, or scoping exercises. However, only a subset of studies explicitly define measurable educational goals and objectives (Step 3), and even fewer describe detailed instructional strategies (Step 4) or implement their proposed curricula (Step 5). Evaluation and feedback mechanisms (Step 6) were rarely reported, and when included, they were typically limited to short-term student feedback or pre-post knowledge assessments. Longitudinal evaluations and outcome-based assessments remain largely absent. The findings underscore a critical implementation gap and emphasize the need for structured, theory-informed, and empirically evaluated AI education models tailored to medical and allied health curricula.

Table 3.

Mapping AI integration in medical education: A comparative analysis using kern’s Six-Step curriculum framework

Study (Author, Year)	Step 1: Problem Identification	Step 2: Needs Assessment	Step 3: Educational Goals & Objectives	Step 4: Instructional Strategies	Step 5: Implementation	Step 6: Evaluation & Feedback
Krive et al., 2023 [13]	Identified a significant gap in undergraduate medical students’ understanding and application of AI in clinical settings.	Assessed through the development and pilot testing of a structured 4-week elective, recognizing low baseline AI literacy and high learner interest.	Defined clear objectives to improve understanding of AI fundamentals, applications in diagnostics, and ethical implications.	Delivered through interactive lectures, clinical case-based discussions, and hands-on data annotation and model testing.	Implemented at a U.S. medical school as an elective module, integrated into the preclinical curriculum.	Evaluated through quiz scores (97%) and assignment scores (89%), with strong student feedback indicating high satisfaction and increased confidence. However, no long-term follow-up was conducted.
Ma et al., 2024 [12]	Identifies the absence of standardized frameworks to guide AI competency development in medical education globally.	Conducted a scoping review of 37 studies and institutional efforts to define existing gaps and recurring components in AI education.	Proposed a Four-Dimensional Literacy Framework encompassing Foundational Knowledge, Application Skills, Ethical Reasoning, and Critical Engagement.	Suggested modular implementation strategies mapped to undergraduate and postgraduate levels, but actual teaching methods were not piloted.	No direct implementation was conducted; the framework is proposed for future curricular adaptation.	No evaluation was performed as the framework remains conceptual. The paper calls for longitudinal assessment post-implementation.
Crotty et al., 2024 [11]	Identified limited integration of AI, particularly machine learning (ML), into radiology and broader medical curricula despite its growing clinical relevance.	Conducted literature review and stakeholder consultation indicating a gap in AI preparedness among medical graduates, especially in interpreting AI-assisted diagnostics.	Aimed to build foundational ML knowledge, foster ethical awareness, and promote critical engagement with AI tools in clinical radiology.	Proposed a modular curriculum consisting of sequential units on ML fundamentals, data bias, ethical use, and clinical integration. Intended to be delivered via lectures, problem-based learning, and simulated cases.	Proposed for piloting in academic settings, but not yet implemented at the time of publication. Discussed collaboration with radiology departments for integration.	Evaluation strategy suggested (e.g., pre/post knowledge assessments), but no actual pilot or feedback data were reported.
Sharma et al., 2023 [19]	Highlighted widespread unawareness and misconceptions about AI among Indian medical students.	Conducted a national cross-sectional survey involving 1,486 students from 243 institutions to assess perceptions, knowledge gaps, and willingness to adopt AI in learning.	Suggested goals for improving AI familiarity, ethical awareness, and readiness to use AI tools in clinical learning.	Recommended integration of AI education through lectures, workshops, online modules, and practical exposure to data tools.	No pilot implementation was conducted; the paper functions as a position statement based on survey findings.	No outcomes or feedback mechanisms were tested, as the study remained exploratory in nature.
Mousavi et al., [21]	Identified a lack of AI and data science preparedness among UK medical students despite increasing clinical use of AI.	Conducted a nationwide survey (n = 525) among medical students to explore current knowledge, confidence, and interest in AI topics.	Called for AI-related competencies to be embedded early, focusing on AI literacy, critical appraisal, and ethical reasoning.	Suggested embedding AI content into existing curricula using a phased, spiral model, integrating lectures, online modules, and case-based learning.	No curriculum was piloted; the paper is recommendation-based. Authors emphasize that integration should be guided by national standards.	Not performed; future research and piloting were recommended to assess impact and feasibility.
Zheng et al., [22]	Identified insufficient AI exposure in Chinese medical curricula despite national mandates for healthcare innovation and AI integration.	Conducted cross-sectional survey among 2,049 medical students to assess perceptions, current knowledge, willingness to learn, and attitudes toward AI in medicine.	Recommended developing competencies in data literacy, critical thinking, clinical application of AI, and ethical judgment.	Suggested development of core modules covering AI principles, clinical applications, ethics, and hands-on data practice—preferably in an interdisciplinary format.	No curriculum piloted; the study makes evidence-based recommendations for curriculum reform in China.	Not applicable—implementation not conducted. The study recommends future outcome evaluations post-implementation.
Rincón et al., [23]	Addressed the critical gap between rapid AI development and underpreparedness of medical graduates to apply AI safely and ethically.	Discussed the misalignment between current curricula and competencies needed for digital transformation in healthcare, drawing from global education policy reviews.	Proposed core competencies including AI ethics, data handling, patient-centered AI use, and interdisciplinary collaboration.	Advocated for integration of AI across the undergraduate curriculum using active learning, interdisciplinary teaching, and formative assessments.	Descriptive and conceptual—no direct implementation reported, but models and guidelines were developed to assist institutions.	Not conducted; called for development of global outcome metrics and suggested use of competency-based evaluations post-integration.
Wood et al., 2021 [14]	Recognized the growing presence of AI in healthcare and the lack of formal AI instruction in U.S. medical curricula. The study highlighted the mismatch between technological advancement and medical education preparedness.	Conducted a cross-sectional survey among 117 students and 44 faculty to explore knowledge, attitudes, and interest in integrating AI into medical education. Revealed gaps in understanding, with many learning about AI via media rather than formal teaching.	Suggested developing AI literacy through curricular components that focus on patient care applications, digital imaging, genomics, and clinical decision support systems. Emphasized interdisciplinary understanding and ethical considerations.	Proposed integration of AI topics longitudinally into existing courses, development of electives and research opportunities, and use of multidisciplinary faculty teams. Also recommended training educators to teach AI.	No formal curriculum implemented; recommendations were made based on survey findings. The study proposed directions for longitudinal integration and development of extracurricular pathways.	No program was implemented at the time of study; evaluation was not applicable. The authors noted that future curricular interventions should include formal outcome assessments.
Grunhut et al., [24]	Recognized the urgency to train future physicians in AI due to its growing influence in healthcare, noting that current curricula globally are underprepared.	Conducted a scoping review of 70 articles and a stakeholder consultation to identify AI-related educational gaps. The review revealed the absence of structured frameworks for implementation and evaluation of AI training.	Developed an AI in Medical Education (AIME) framework with four pillars: foundational knowledge, clinical application, professionalism/ethics, and systems-based practice. This served as the basis for defining AI literacy goals.	Recommended embedding AI literacy longitudinally across the medical curriculum using case-based learning, interdisciplinary modules, and technology-enhanced pedagogy. Also highlighted the use of flipped classrooms and faculty development.	Proposed strategies and implementation roadmaps through the AIME framework but did not document a full curriculum roll-out. Pilots and exemplar modules were recommended.	Acknowledged the lack of outcome-based evaluations in existing studies and stressed the need for future research to validate effectiveness and sustainability. Suggested incorporation of Kirkpatrick’s model for future evaluations.
Naseer et al., [25]	Identified a growing gap between the evolving role of AI in healthcare and the absence of AI-related education in current medical curricula, particularly in Pakistan and similar LMIC contexts.	Conducted a survey among 309 undergraduate medical students at a public-sector medical university in Karachi, Pakistan. Assessed awareness, understanding, and attitudes toward AI. Results revealed a significant knowledge gap and enthusiasm for curriculum integration.	Advocated for introducing AI literacy early in medical training, with objectives focusing on demystifying AI, understanding its limitations, and exploring clinical relevance.	Proposed integrating AI into existing curricula through seminars, workshops, and elective modules. Emphasized the need for interdisciplinary collaboration, especially involving computer science departments.	Suggested a phased and low-cost integration of AI components using available institutional infrastructure, particularly in resource-limited settings. No implemented curriculum was reported at the time of study.	Not implemented, but the study emphasized the need for longitudinal monitoring of AI education outcomes and feedback-driven curriculum refinement.
Mondal & Mondal, 2024 [17]	The study identifies a growing concern regarding the ethical implications of AI integration in healthcare, especially concerning informed consent, data privacy, and algorithmic bias.	The authors emphasize gaps in ethical preparedness among healthcare professionals, especially in developing countries, where informed consent and equitable access are often overlooked.	The implicit educational goal is to enhance awareness and critical understanding of ethical AI application, including informed consent, bias mitigation, and responsible decision-making.	Five case studies and a real-world narrative are used to deliver ethical dilemmas and contextual learning. While not structured as formal curriculum modules, these are instructional tools intended to provoke reflection and applied learning.	Not implemented as a formal intervention. However, the chapter proposes integrating ethical considerations of AI into broader healthcare and microbiology education.	No structured evaluation or assessment methodology was reported. The emphasis is on narrative learning rather than measurable outcomes.
Weidmann et al., [26]	Highlights the insufficient integration of AI in pharmacy education globally, with a lack of clarity on competencies and curriculum standards needed to prepare pharmacists for AI-enhanced healthcare systems.	Based on a scoping review of 29 studies, the authors identify that AI training in pharmacy is fragmented and primarily exploratory, indicating a need for structured curriculum development.	Recommends defining core AI competencies for pharmacy students, including understanding AI terminology, ethical implications, clinical applications, and critical thinking for AI-informed decision-making.	Suggests using interdisciplinary teaching, simulation tools, problem-based learning, and online modules to teach AI principles. Some reviewed studies reported using didactic lectures and short courses.	Not implemented uniformly; the paper calls for a global collaborative approach for AI integration and mentions isolated cases of elective modules being offered in Canada, US, and UK.	Notes that most studies lack robust assessment strategies; evaluation is often self-reported by students. Authors advocate for future research to include objective outcome evaluation.
Tolentino et al., [27]	Identifies a growing demand for AI literacy in medicine and the lack of a standardized framework to guide AI curriculum development in medical education.	The authors conduct a scoping review and engage stakeholders from medicine and computer science to assess gaps in current curricula and desired competencies.	Proposes a conceptual curriculum framework with three domains: (1) Data Literacy, (2) Algorithmic Literacy, and (3) Ethical-Legal-Social Implications (ELSI). Objectives include critical understanding of AI tools, bias identification, and ethical reasoning.	Recommends interdisciplinary team-teaching, flipped classrooms, and problem-based learning methods, though these remain theoretical suggestions.	The framework is presented as a proposed model but not yet implemented. It serves as a guide for future curricular integration across undergraduate and graduate levels.	No specific evaluation tools or outcome measures are reported; future studies are encouraged to pilot and validate the framework.
Blanco et al., [28]	Recognizes that despite the rise of AI in healthcare, AI integration into medical education is minimal, leaving graduates underprepared for future clinical practice.	Based on a narrative synthesis of published literature, the authors identify a lack of structured AI competencies, curricula, and faculty development as major barriers.	Suggests that medical education should include training in AI fundamentals, critical appraisal of AI tools, and ethical reasoning. However, explicit learning objectives are not detailed in a curriculum.	Highlights examples from other studies using flipped classrooms, problem-based learning, simulation, and case-based discussions. Recommends interdisciplinary teaching approaches.	Reports limited implementation globally; most examples were in pilot or elective stages, particularly in North America and Europe.	Notes that most studies lack robust evaluation; calls for longitudinal assessment strategies to evaluate competency development and clinical application.

This conceptual model is informed by thematic synthesis and integrates principles from existing frameworks (FACETS, TAM, Diffusion of Innovation) while aligning with Kern’s six-step approach for curriculum design.

Discussion

This review highlights the extensive use of AI in medical education, spanning from diagnostic support to adaptive learning platforms. However, most initiatives remain at pilot or conceptual stages, with limited outcome data available to guide best practices. Many AI tools employed in training still lack thorough validation and have yet to be integrated into standard curricula. The comparative synthesis further highlighted a geopolitical divide in AI education readiness. Institutions in high-income countries reported greater access to digital infrastructure, interdisciplinary collaboration, and implementation support, enabling more seamless AI integration. In contrast, institutions in low- and middle-income countries (LMICs) faced barriers such as limited AI literacy among faculty, budget constraints, and lack of localized AI tools. These disparities emphasize the need for adaptable AI education strategies, such as open-access resources and context-sensitive faculty development, to avoid exacerbating global educational inequities. To enhance the operational utility of our review, we mapped selected studies to Kern’s Six-Step Curriculum Development Model (Appendix A). This mapping highlights that while most studies addressed early stages such as problem identification and instructional design, very few incorporated structured implementation or outcome evaluation. The limited attention to Steps 5 and 6 underscores a critical gap in curriculum sustainability and impact assessment.

Curriculum development remains a significant hurdle. Although models such as FACETS [29], the four-dimensional AI literacy framework [12], and modular designs [13] have been proposed, few have undergone real-world testing or long-term evaluation. A gap persists between theoretical enthusiasm and practical application, often compounded by institutional resistance, funding shortages, and a scarcity of faculty skilled in AI. Applying Kern’s model could help bridge this divide by focusing on learner needs, guiding rollout, and systematically assessing educational outcomes [30].

Ethical concerns add another layer of complexity. As AI increasingly informs clinical decision-making, students require more than technical skills—they must also critically evaluate algorithmic outputs. The review emphasizes growing demands for education on bias reduction, algorithmic transparency, and data governance, yet these topics are seldom embedded within core curricula, often remaining fragmented or optional. Tools like explainable AI (XAI) [31] and frameworks such as TRIPOD-AI [32] offer ways to connect technical details with ethical understanding.

Engaging key stakeholders is crucial for successful implementation. Students demonstrate keen interest in AI training, especially in clinical reasoning and diagnostics. Faculty support exists but is frequently tempered by limited expertise and confidence. Interdisciplinary collaboration among educators, clinicians, data scientists, and ethicists is vital for creating effective curricula. Initiatives like AI hackathons, cross-disciplinary electives, and co-teaching programs provide practical avenues to foster this cooperation [33].

Infrastructure and equity concerns must also be addressed. Many AI tools demand significant technological resources, which may be beyond reach for institutions in low-resource settings. Without coordinated planning and international sharing of resources, AI education risks deepening existing disparities. Prioritizing open-access tools and scalable, resource-light strategies will be essential [34].

Another critical barrier to effective AI integration is the shortage of faculty with adequate expertise in AI principles and applications. To address this, structured faculty development models should be prioritized. These could include micro-credentialing programs offering modular, competency-based certification in AI literacy, as well as collaborative bootcamps co-developed with computer science departments to provide interdisciplinary, hands-on training. Investing in such faculty development strategies is essential to ensure sustainable and contextually relevant delivery of AI curricula.

A major strength of this systematic review is its comprehensive and methodologically rigorous approach. We adhered to the PRISMA 2020 guidelines and employed a transparent protocol that included searches across multiple databases (PubMed, Scopus, Web of Science, and Google Scholar), supplemented by manual screening of references and grey literature. We applied a structured thematic synthesis and used validated tools (ROBINS-I, NOS, CASP, AMSTAR 2, AXIS) to assess the risk of bias across diverse study designs, ensuring methodological integrity.

However, several limitations should be noted. First, the heterogeneity in study designs, outcomes, and reporting limited the feasibility of conducting a meta-analysis. Second, the included studies were predominantly from high-income countries, restricting the generalizability of findings to low- and middle-income settings (LMICs), where educational infrastructure, curricular demands, and access to AI tools may vary significantly. Third, many studies lacked rigorous evaluation frameworks and relied heavily on subjective, short-term outcomes such as self-reported knowledge or attitudes, which are prone to recall and social desirability bias. Notably, few studies assessed the long-term impact of AI interventions on learner performance, clinical competence, or patient outcomes. These gaps underscore the urgent need for more longitudinal, outcome-based research to evaluate the sustained effectiveness and real-world applicability of AI integration in medical education.

Conclusion

AI is reshaping medical education by enhancing teaching methods, increasing assessment accuracy, and enabling personalized learning experiences. This review underscores the importance of structured, ethically grounded, and evidence-based AI training programs. A phased integration strategy—starting with foundational literacy during preclinical education and advancing toward more complex applications in clinical training—offers a practical and sustainable pathway. As AI technologies evolve, educational systems must adapt accordingly, equipping future clinicians not only to use these tools proficiently but also to lead thoughtfully and ethically in AI-driven healthcare.

Abbreviations

AI

Artificial Intelligence

AXIS

Appraisal tool for Cross-Sectional Studies

CASP

Critical Appraisal Skills Programme

JBI

Joanna Briggs Institute

NOS

Newcastle-Ottawa Scale

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

ROBINS-I

Risk Of Bias In Non-randomized Studies - of Interventions

SANRA

Scale for the Assessment of Narrative Review Articles

Authors’ contributions

ZA conceptualized the study, conducted the literature search and data extraction, analyzed and synthesized the findings, and wrote the manuscript. All phases of the review were conducted independently and verified by at least one other reviewer. The author has read and approved the final manuscript.

Funding

None.

Data availability

All data analyzed during this study are included from the published article and its supplementary information files. The full list of included studies and extracted data is available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.