Mapping the use of artificial intelligence in medical education: a scoping review | BMC Medical Education

This scoping review provides an comprehensive overview of the integration of artificial intelligence in UME, highlighting the diversity of approaches, ethical considerations, and global discrepancies. The review also identifies critical research gaps and limitations while proposing recommendations for enhancing AI curricula to meet the evolving needs of healthcare.

Diversity in AI curriculum content and delivery

The studies reviewed demonstrated considerable variability in AI curriculum development and delivery. Some focused on foundational AI concepts, such as machine learning and deep learning, introduced during pre-clinical years, while others employed a staged approach, integrating these concepts in clinical contexts [41, 45,46,47, 54]. This heterogeneity underscores the lack of a standardized framework, complicating the creation of cohesive and comprehensive AI education models that align with modern healthcare demands. As highlighted in the results described in 3.2, collaborative learning, teamwork skills, interdisciplinary abilities, and basic digital skills are emerging as transversal competencies. These competencies provide curricular support for integrating AI as a tool rather than as a subject of study itself, facilitating its adoption across diverse educational settings without necessarily establishing it as standalone content. Establishing consensus on essential AI skills and competencies is necessary for ensuring consistency and effectiveness in AI integration across UME programs.

Ethical education and digital competence

Ethical training and digital literacy emerged as fundamental components in AI curricula. Many studies stressed the importance of preparing students to navigate ethical challenges related to AI, such as maintaining patient confidentiality, recognizing algorithmic biases, and ensuring informed consent [28,29,30,31,32]. However, despite this consensus, the literature lacks comprehensive strategies and standardized curricula to systematically teach these competencies. This gap indicates a need for further curriculum development to incorporate structured and evidence-based approaches to ethical education.

Existing studies have demonstrated that AI applications, such as Intelligent Tutoring Systems and virtual simulations, significantly enhance learning outcomes by providing adaptive, personalized training and reducing geographical barriers to education [23, 24, 53]. These tools have been shown to improve self-directed learning and critical thinking skills, essential for navigating AI-integrated clinical environments [24, 25]. Early research on learning outcomes also highlights that a staged integration of foundational AI concepts, followed by their application in clinical settings, fosters better knowledge retention and practical skill development [29, 32, 51].

Faculty readiness is a critical factor in the successful integration of AI into medical education. Many medical educators lack formal training in AI, which can hinder effective curriculum implementation. Providing faculty development programs, workshops, and interdisciplinary collaborations with AI specialists can bridge this gap. Furthermore, AI literacy training should focus not only on technical aspects but also on ethical considerations, clinical applications, and teaching strategies. Without adequate faculty support, AI education risks being inconsistently implemented, leaving students without the necessary guidance to critically engage with AI-driven tools in their future clinical practice.

By addressing these gaps through standardized curricula, medical education can better prepare future physicians to responsibly and effectively use AI technologies while upholding the ethical principles critical to patient care.

Pedagogical approaches and experiential learning

A variety of pedagogical methods were proposed across the studies, including lectures, e-learning modules, collaborative learning platforms, and hands-on experiences with AI tools [46, 47, 50, 54,55,56]. Experiential learning was consistently recommended as a critical component for effective AI training; however, the lack of robust evaluations and evidence-based frameworks limits understanding of the most effective approaches. AI, as a tool for teaching and learning, can be used interactively, allowing students to learn from it without necessarily being the object of study itself. Further research is needed to identify optimal pedagogical strategies and establish structured evaluations to assess student outcomes comprehensively.

The lack of standardization in AI curricula across medical schools presents a major challenge to ensuring consistency in AI education. While some institutions have introduced AI as a standalone subject, others integrate it into existing courses with varying levels of depth. A possible solution is to develop a core competency framework outlining the essential AI-related skills that all medical students should acquire. Collaborative efforts between medical education governing bodies, AI experts, and curriculum developers could help establish standardized AI learning objectives. Additionally, accreditation bodies could play a role in encouraging AI integration by setting minimum competency requirements for medical graduates, ensuring that all students, regardless of institution, receive foundational training in AI applications relevant to clinical practice.

Global discrepancies in AI integration and student perceptions

The geographical diversity of the studies highlights significant global discrepancies in AI integration and student perceptions. While North American and European studies emphasized the need for formal, structured AI training to enhance clinical decision-making, studies from regions like Nigeria and India expressed concerns about the dehumanization of healthcare and the risk of technology overreliance [33,34,35,36,37]. Additionally, North American and European models often integrated interdisciplinary approaches, fostering collaboration between medical and technological disciplines to enhance critical thinking and teamwork skills [23,24,25,26,27]. In contrast, studies in regions like Nigeria highlighted a limited exposure to interdisciplinary applications, focusing more on standalone AI tools as solutions to immediate challenges [33,34,35,36,37]. These regional differences suggest that, although there is a shared acknowledgment of AI’s importance, curricula must be tailored to cultural and institutional contexts to ensure their relevance and effectiveness globally.

Through the strategic use of AI, countries in development can overcome some challenges and improve the quality and reach of medical training. Among the key approaches, AI enables access to personalized learning materials, using low-cost applications to tailor educational resources to individual levels of knowledge. AI also enables the adaptation of content to local languages and cultural contexts, helping learners better understand information. In addition, chatbots and virtual tutors can offer constant support, resolving doubts and providing feedback in real time, also facilitating virtual clinical simulations and practices, training students in complex situations without the need for physical facilities and at lower cost, while image recognition algorithms support learning in assisted diagnosis, especially useful in areas with few specialists [23, 32, 33]. Despite its advantages, the use of AI in low-resource settings faces challenges, such as limited technological infrastructure, lack of connectivity in rural areas, and the need for technology training for teachers and students. However, if implemented in an adaptable, accessible and sustainable manner, AI could have a positive and lasting impact on medical education in these communities.

These findings underscore the necessity of aligning AI education frameworks with regional healthcare priorities and cultural values. Tailoring interdisciplinary models to fit local contexts ensures that AI integration addresses the diverse needs of medical education systems globally, enhancing both its applicability and impact.

Ethical challenges and opportunities of AI in transforming medical education

The ethical implications of AI integration in medical education are critical for ensuring that the use of AI enhances learning without compromising the human values inherent to healthcare practice. AI lacks the capacity for empathy and moral judgment, which are essential characteristics of health professionals [29]. This raises concerns about the potential impact of increased AI reliance on the development of interpersonal skills and the ability to respond compassionately to patients’ emotional and psychological needs [55]. As AI tools become more prominent, bioethics education must adapt, ensuring that AI is viewed as a complementary tool while emphasizing the healthcare professional’s responsibility in decision-making [48]. Additionally, equity and access are ethical considerations that must be addressed. While AI has the potential to improve medical education globally, unequal access could amplify existing disparities, disadvantaging students and professionals in resource-limited settings [44]. Concerns about privacy and data security are also significant, particularly when using AI tools like chatbots that handle sensitive patient information [40]. Developing strict regulations and ethical frameworks is essential to protect confidentiality and prevent legal issues.

Educational innovation: the impact of AI on teaching and learning in medicine

AI is revolutionizing medical education by providing innovative tools that enhance both teaching and learning [38, 55, 56]. On the instructional side, AI applications are used for plagiarism detection, curriculum development, automated feedback, and clinical simulations [47]. For students, AI serves as a resource for addressing questions, practicing clinical cases in safe environments, and receiving real-time feedback. Perhaps the most promising focus in medical education is precision medical education, similar to precision medicine, which allows for personalized competency development in students. AI, with its foundational capabilities in analytics, supports this approach, enabling tailored learning experiences [46]. In this regard, and contrasting with geographical variability, there is a highlighted need for standardization and adaptation of foundational skills necessary for AI adoption, continual development, and innovation within medical education. These tools not only promote interactive and efficient education but also prepare future physicians to navigate technological changes and the globalized healthcare landscape.

Systemic and regional barriers to AI curriculum adoption

Despite AI’s transformative potential in medical education, its integration remains uneven across regions, particularly in low-resource environments. Systemic barriers—ranging from infrastructure limitations to faculty readiness and regulatory gaps—vary significantly depending on local economic and technological contexts.

Financial and infrastructure constraints

Medical schools in low- and middle-income countries (LMICs) often lack access to AI-driven educational tools, including high-speed internet, cloud computing, and AI-based simulation platforms. Institutions in high-income countries, by contrast, benefit from substantial investments in AI-enhanced learning environments. AI implementation in LMICs is hindered by limited digital infrastructure and the high costs of AI integration, restricting access to advanced AI-powered educational tools and training programs ​ [57]. Cloud computing has been proposed as a potential solution, allowing resource-limited institutions to access AI tools without the need for significant in-house infrastructure investments​ [57].

Faculty training deficits

Faculty AI literacy varies across regions, with educators in resource-limited settings often having limited access to AI training programs. Without structured faculty development, AI integration remains inconsistent. One of the major barriers identified is the scarcity of AI-related instructional resources tailored to LMICs, which restricts faculty engagement with AI-based medical education [58]. Expanding access to faculty training programs and leveraging online AI education platforms could help mitigate these barriers.

Ethical and regulatory uncertainty

Many institutions face unclear policies regarding AI in medical education, particularly concerning data privacy, AI-assisted assessments, and ethical guidelines. In LMICs, AI adoption in healthcare is further challenged by a lack of regulatory frameworks ensuring AI tools align with local health systems and ethical standards, leading to inconsistencies in implementation​ [57]. Additionally, the limited availability of high-quality training datasets specific to LMICs creates further challenges for AI development and deployment, as most AI models are trained on data from high-income countries, potentially introducing bias and reducing local applicability​ [58].

Study limitations and research justification

Limitations in methodological rigor and generalizability

Many studies in this review exhibit limitations in methodological rigor, including the absence of standardized evaluation metrics and reliance on self-reported data from small, non-representative samples, which affects the strength and generalizability of findings. Additionally, the scarcity of longitudinal studies impedes a comprehensive understanding of AI training’s long-term impact on clinical decision-making and skill transfer. Addressing these methodological limitations through multi-institutional studies with robust designs is essential for creating reliable, scalable models for AI education.

Research gaps in low-resource contexts

A critical gap identified is the limited exploration of AI integration in medical education within low-resource settings. Investigating how AI curricula can be adapted to regions with limited technological infrastructure, while still fostering essential competencies, presents a valuable research opportunity. Future studies should prioritize this area to develop adaptable and accessible AI education models that promote equitable training standards globally.

Need for standardized pedagogical approaches

The lack of standardized pedagogical approaches in AI curricula hinders the development of cohesive training frameworks. Further research is required to examine how standardized curricula influence clinical skills and decision-making. Multi-institutional, diverse studies can address this gap, providing insights into the effectiveness of different educational models and informing evidence-based best practices for AI education in medical contexts.

This scoping review acknowledges potential biases in the study selection and analysis processes. Limiting the inclusion criteria to studies published in English and Spanish may have excluded relevant research in other languages, reducing the diversity of perspectives. Similarly, the reliance on specific databases like PubMed, Scopus, and BIREME might have introduced selection bias by underrepresenting certain regions or less-accessible research.

Potential bias in study selection and analysis

This review excluded conference abstracts and grey literature, which may contain early-stage findings, pilot programs, and evolving technological advancements in AI applications for medical education. While these sources can provide valuable insights, they often lack peer review, standardized methodologies, and comprehensive data, making it challenging to assess their rigor and reproducibility. Additionally, the thematic analysis method, while systematic, involves subjective interpretation, which may have unintentionally emphasized certain findings over others. The iterative refinement of inclusion criteria during the study selection process may also have influenced the final set of included studies, potentially impacting the comprehensiveness of the results.

These limitations highlight the need for caution when generalizing findings, particularly for underrepresented regions and contexts. Future research should consider expanding the linguistic and geographical scope, incorporating select grey literature, and employing independent validation methods to enhance the reliability and applicability of results.

Recommendations for future research and practice

Create a standardized, consensus-based AI curriculum with defined competencies

Developing a standardized AI curriculum requires actionable steps to ensure its effective implementation. Curriculum developers should organize interdisciplinary workshops involving educators, AI experts, and healthcare professionals to identify core competencies. To achieve this, working groups should be established, bringing together medical educators, AI specialists, and accreditation bodies to define the core AI competencies required for medical students. Using the Delphi method can help reach consensus on minimum AI-related learning objectives, ensuring alignment with accreditation standards. Additionally, pilot programs should be introduced across multiple institutions, with data collection on student performance to evaluate effectiveness before large-scale implementation.

Embed ethical training as a core component alongside practical skills development

Embedding ethical training into AI-related courses can be achieved by designing case-based learning modules that address real-world challenges such as algorithmic bias, patient privacy, and decision accountability. Interdisciplinary collaborations between medical faculties and bioethics departments should be established to develop structured ethical AI coursework. Furthermore, mandatory assessments should be incorporated to evaluate students’ ability to navigate ethical AI applications in clinical practice. These measures will help ensure that students not only understand AI’s technical capabilities but also develop the ethical reasoning necessary to apply AI responsibly in healthcare.

Implement experiential learning modules for applied AI skills

Experiential learning should focus on simulation-based activities where students engage with AI tools like diagnostic algorithms and clinical decision support systems. Introducing AI-powered clinical simulations will allow students to interact with AI-driven decision-support tools in a controlled environment, helping them understand their potential and limitations. Additionally, AI-focused problem-based learning (PBL) sessions should be incorporated, where students analyze real medical cases using AI-generated insights. Partnering with health technology companies will provide students with hands-on exposure to AI applications in medical practice, further reinforcing their practical skills.

Conduct research and evaluation to validate curriculum impact on clinical competencies

Validating the impact of AI curricula involves implementing pilot programs to test their effectiveness. Institutions should collect baseline data and track outcomes such as clinical competency, decision-making skills, and patient care quality. Longitudinal studies should be conducted to assess the long-term impact of AI education on medical students’diagnostic accuracy and clinical reasoning. Standardized evaluation metrics should be developed to compare AI curricula across different institutions, ensuring a comprehensive understanding of best practices. Findings from these studies should be disseminated through international AI in medical education conferences to promote continuous improvements and knowledge sharing.

Adapt curriculum to regional needs and resource availability

Adapting AI curricula to regional contexts requires conducting needs assessments to identify barriers, such as limited infrastructure or access to AI tools. Flexible curriculum models should be developed, offering both high-tech AI learning tools and low-resource alternatives, such as AI case studies and offline learning modules for regions with limited digital access. Additionally, partnerships with governmental and non-governmental organizations should be encouraged to provide funding and support for AI curriculum implementation, ensuring inclusivity and relevance across diverse educational settings.