Policy Brief: AI Readiness in Medical Education: Assessing Current Gaps and Future Outlook

The policy problem

The integration of artificial intelligence (AI) and related technologies into various aspects of human life has become increasingly prevalent, significantly influencing the healthcare sector.¹ AI technologies have developed sophisticated algorithms to analyze diverse health data, including clinical, behavioral, environmental, and pharmaceutical information, by leveraging patient data and biomedical literature.² However, the incorporation of these technologies into medical practice requires a workforce skilled in the technical, ethical, and practical aspects of AI.³ With enhanced access to health data and substantial investments by technology companies in AI, its applications in medicine are becoming increasingly valuable. For example, AI systems now assist healthcare professionals in fields such as radiology, pathology, and precision oncology.⁴ Moreover, AI is instrumental in improving patient care through innovations such as remote patient monitoring, telemedicine, and virtual support systems.⁵

Artificial Intelligence (AI) is a prominent and rapidly evolving topic within technological advancements,⁶ with considerable potential to impact the healthcare industry, particularly in medical education. AI has the capacity to transform medical education by providing personalized and adaptive learning experiences, enhancing diagnostic accuracy, and facilitating data-driven decision-making processes.⁷ In contrast to traditional approaches that often employ a uniform, rote-learning model for all students, AI allows for the customization of learning processes to meet individual needs, enabling students to concentrate on areas requiring further practice.⁸ Modern medicine generally adopts a forward-looking perspective toward these identified challenges. This future-oriented approach enhances the appeal of AI applications in healthcare, which appear increasingly integrated into the medical field.⁹ As noted, futurist Eric Topol asserts, “Virtually every physician in the future—from specialists to paramedics—will utilize AI technologies, particularly deep learning.” This statement underscores the extensive scope of AI’s application in medicine.² Furthermore, AI can assist educators in designing individualized curricula, continuously monitoring learners’ progress, and providing immediate feedback. Although prior studies have primarily focused on examining specific types of AI and their effectiveness in medical education.¹⁰

Numerous studies have examined the essential knowledge that medical students should gain concerning artificial intelligence (AI) in medicine. Additionally, some research has highlighted the importance of integrating health AI ethics education into medical school curricula. Students view AI as a promising enhancement to the future of medicine and argue that it should be considered a collaborator rather than a competitor. Furthermore, they believe that training in AI can significantly impact their career trajectories.^11,12

Emerging skills such as prompt engineering— the art of crafting precise inputs for large language models (LLMs) like ChatGPT— are critical for future physicians. Prompt engineering can enhance medical education by generating realistic patient scenarios, multiple-choice questions for assessments, or personalized explanations of complex concepts, thereby bridging the gap between theoretical knowledge and practical application. For instance, in decision support systems, well-engineered prompts can optimize AI for accurate diagnosis and ethical considerations, reducing algorithmic biases. However, current curricula often overlook this skill, leading to suboptimal AI utilization in healthcare.^13–15

Despite these advancements, the implementation of artificial intelligence (AI) education in medical schools globally remains inconsistent. Key deficiencies include the absence of standardized curricula, inadequate practical application of AI tools, a shortage of faculty with expertise in AI, limited knowledge and skills among students, and varied attitudes and levels of preparedness among learners.^1,16 Moreover, critical aspects such as the ethical implications and policy considerations of AI are often insufficiently addressed in current educational programs.^17,18 These deficiencies underscore the urgent need for the development of comprehensive and standardized educational frameworks to adequately prepare future physicians for integrating AI into healthcare systems. Medical students, as a vital stakeholder group, are central to discussions about the future of healthcare, and their perspectives on AI applications are significant. Research indicates that, in many instances, medical students believe they understand the concept of AI; however, when asked to define it, the majority are unable to do so.^11,12 The existing literature emphasizes the necessity of incorporating AI application training into medical curricula, highlighting that current education in this area is neither sufficient nor satisfactory.^11,12,19 Although students anticipate that AI will transform and revolutionize healthcare, they recognize that current training in this domain is inadequate.¹¹ The objective of this policy brief is to establish a framework for AI readiness in medical education, assess current gaps, and evaluate the future outlook for AI in healthcare. This framework will assist policymakers in universities, governments, and health authorities in designing effective educational programs.

The study

This policy brief presents the findings of a study conducted in 2024 involving 1,916 medical students from years one to five at kermanshah University of Medical Sciences. The study utilized a census sampling method. The instrument employed was the Medical Artificial Intelligence Readiness Scale for Medical Students (MAIRS-MS), developed by Karaca et al., which consists of 22 questions across four subscales: cognition, ability, vision, and ethics. All participants provided written informed consent before completing the questionnaire. The consent process included comprehensive information regarding the study’s purpose, procedures, potential risks and benefits, and the voluntary nature of participation. Participants were informed of their right to withdraw at any time without facing any consequences. Since all participants were medical students aged 18 years or older, no minors were involved, and therefore, parental consent or assent was not required. Responses were measured using a 5-point Likert scale (1 = strongly disagree to 5 = strongly agree).²⁰ The overall mean score for the scale in this study was 3.45 ± 0.40. Among the subscales, “vision” received the highest score, while “cognition” received the lowest ( Figure 1). The validity and reliability of the Persian version of this questionnaire were previously established by Ghalibaf et al. (2023) among medical students at Mashhad University of Medical Sciences, with Cronbach’s alpha coefficients of 0.886, 0.905, 0.865, and 0.856 for the cognition, ability, vision, and ethics subscales, respectively, and an overall Cronbach’s alpha of 0.944 for the entire scale.²¹ Similarly, a study by Rezazadeh et al. among medical students in Kerman confirmed the questionnaire’s validity and reliability, reporting an overall Cronbach’s alpha of 0.94.²² In the present study, the Cronbach’s alpha coefficient was 0.762.

Figure 1. The mean of the artificial intelligence variables of medical students.

The findings indicate that the mean medical artificial intelligence readiness score was 3.45, with a standard deviation of 0.40, suggesting a moderate level of preparedness. Building upon the original findings (mean: 3.45 ± 0.40), a larger cohort analysis conducted in 2025 reaffirms moderate readiness (61.34 ± 10.13). ANOVA results indicate significant differences across academic years (p < 0.001), with third-year students exhibiting the highest readiness levels at 64.85 ± 8.01. Additionally, regression analysis identifies prior exposure to AI as a predictor of cognition and ability (β = 0.092, p = 0.020; β = 0.113, p = 0.004, respectively). Gender differences were found to be non-significant for AI subscales (p > 0.05). The highest score in the vision subscale reflects optimism regarding AI’s potential to enhance diagnostic accuracy and improve patient outcomes. Scores in the ability and ethics subscales demonstrate moderate confidence in the utilization of AI tools, along with an awareness of the ethical challenges associated with them. Conversely, cognition was identified as the weakest subscale, highlighting a limited understanding of AI concepts such as machine learning. The variability in scores suggests disparities in access to AI education. These findings are consistent with global trends, as only 10% of medical schools worldwide have integrated AI into their curricula.²³

Discussion

The moderate readiness score of 3.45 is consistent with global findings, highlighting the necessity for targeted interventions in cognitive skills, where prompt engineering could be instrumental. By equipping students with the ability to formulate effective prompts, educators can enhance AI literacy, facilitating the improved integration of large language models (LLMs) in clinical practice. This strategy not only improves diagnostic accuracy but also tackles ethical challenges, including bias mitigation through optimized inputs. Incorporating prompt engineering into medical curricula has the potential to transform medical education, rendering it more adaptive and personalized.

Author contributions

AZ conceptualized and designed the survey, conducted the investigation, analyzed the data, revised the manuscript, and performed grammatical editing. AZ has reviewed and approved the final manuscript.

Ethics and consent

The data collection in the present study was conducted after the approval of Kermanshah University of Medical Sciences, and Publication Ethics Boards the number IR.KUMS.REC.1402.472. We confirm that all methods used in this study were carried out in accordance with relevant guidelines and regulations. The participation of students was completely voluntary and informed consent was obtained from all participants.