Ahmad et al [6] introduced the notion of “AI syndrome,” cautioning that reliance on AIGC may result in “progressive cognitive decline” among medical students. The authors argue that students tend to favor the immediate responses offered by AIGC, which diminishes their capacity for independent thought and undermines the development of critical clinical reasoning skills. While these short-term efficiencies may be appealing, they could ultimately contribute to a deterioration of cognitive abilities, thereby jeopardizing clinical judgment and patient safety in the long run.

In 2023, Jeyaraman et al [7] highlighted that while AIGC demonstrates proficiency in memory-related tasks, it exhibits significant shortcomings in deep reasoning, challenging established knowledge and mitigating unwarranted self-confidence. These deficiencies may inadvertently promote a lack of diligence among students, leading them to neglect critical thinking. Concurrently, Prober and Desai [8] advocated for a transformation in medical education, suggesting a shift from prioritizing students with strong memorization skills to fostering essential competencies, such as critical thinking and empathy. This recommendation is underscored by the fact that AI has successfully passed the United States Medical Licensing Examination, indicating a diminishing emphasis on rote memorization. Subsequently, a mixed methods study conducted by Magalhães Araujo et al [9] in 2024 revealed that approximately 20% of students exhibited signs of cognitive inertia, relying heavily on cue words, thereby underscoring the potential adverse effects of AIGC on students’ independent thinking capabilities. In the same year, Xu et al [10] explicitly cautioned against the dangers of excessive dependence on AIGC, warning that it could foster “intellectual laziness.” They noted that the authoritative presentation of information generated by AIGC might lead students to accept such information uncritically, thereby undermining their critical thinking skills. In 2025, Madrid et al [11] further corroborated the limitations of AIGC regarding deep reasoning and the challenge of knowledge in their analysis of the German medical licensing examinations, emphasizing its detrimental effects on students’ cognitive engagement.

In 2023, Jeyaraman et al [7] were pioneers in identifying the potential challenges associated with AIGC within the realm of medical education, advocating for the creation of detection tools to mitigate the risks of AIGC misuse. Following this, a cross-sectional study conducted by Kaliterna et al [12] in 2024 further elucidated the gravity of the issue, revealing that as many as 34% of reflective papers authored by medical students were likely generated or altered by AI. This occurrence not only underscores the significant threat posed to academic integrity by the misuse of AIGC but also indicates a potential detrimental impact on students’ independent thinking and writing skills.

These investigations illustrate that while AIGC offers notable advantages in the retrieval and application of memory-based knowledge, thereby providing efficient and convenient support for medical education, it also presents considerable limitations in fostering higher-order critical thinking. As visualized in Figure 4, AIGC’s capabilities are predominantly situated at the lower levels of the cognitive hierarchy (eg, information retrieval and knowledge comprehension), with diminishing efficacy at higher levels requiring analysis, evaluation, and creation. This hierarchy effectively illustrates the inherent limitations of current AIGC tools in cultivating advanced critical thinking skills. Addressing these challenges necessitates a collaborative effort among educators, students, and policymakers to ensure that the integration of AIGC in medical education enhances, rather than impedes, the comprehensive development of students through appropriate guidance and effective regulation.

Numerous surveys indicate a nuanced and often contradictory perspective regarding AIGC. In a 2024 survey conducted by Abdullah et al [13], it was reported that 69% of medical students acknowledged the potential advantages of AIGC in fostering critical thinking, while 31% expressed concerns regarding its potential to induce mental inertia. A subsequent study by Moskovich and Rozani [14] in 2025 revealed a prevalent apprehension among medical students that AIGC may undermine critical thinking skills and promote complacency. Notably, a 2023 investigation by Almazrou et al [15] indicated that as many as 92.6% of medical students recognized the prospective benefits of AIGC in enhancing critical thinking. Furthermore, research by Abouammoh et al [16] suggested that students exhibited a more pragmatic outlook compared to educators, who tended to highlight “critical thinking deficits.”

These findings imply that while medical students generally perceive AIGC as a valuable tool for improving critical thinking, they simultaneously harbor concerns about the potential for mental inertia, thereby illustrating the dual opportunities and challenges presented by the integration of AIGC in medical education.

In recent years, the utilization of virtual standardized patients (SP) has emerged as an innovative tool in medical education, garnering increasing attention for its potential to enhance the critical thinking skills of medical students. A study conducted by Jiang et al [17] in 2024 revealed that the integration of virtual SP training with case-based learning significantly enhances students’ systematic and evidence-based thinking abilities. However, the study noted that the short-term effects on higher-order critical thinking skills were not as pronounced, indicating that a more profound integration of virtual SP training is necessary to effectively foster the development of critical thinking among medical students. Furthermore, in 2025, Öncü et al [18] highlighted the considerable advantages of AIGC, such as ChatGPT-4o, in terms of cost-effectiveness, accessibility, and standardization as a virtual standardized patient. This advancement presents new opportunities for the optimization and promotion of virtual SP training.

Nevertheless, from the perspective of critical thinking development, further investigation is required to explore how the benefits of AIGC can address the limitations of virtual SP training in enhancing higher-order thinking skills.

In 2024, Shalong et al [19] conducted a randomized controlled trial that revealed the efficacy of a tailored GPT AI tool, which offers structured guidance rather than direct responses, in enhancing students’ scores on the Cornell Critical Thinking Test over a 12-week period and beyond. This finding indicates that the development of critical thinking skills among medical students can be effectively facilitated through the thoughtful design of AIGC guidance tools. In 2025, Reinke et al [20] introduced a framework for “guided use,” which positions AIGC as a critical thinking “partner” rather than a mere “substitute.” Additionally, Elabd et al [21] employed AIGC, such as ChatGPT and DALL-E 3, to create personalized multimodal memory methods tailored to various learning styles, while emphasizing the importance of treating these tools as supplementary resources to mitigate the risk of over-reliance. Furthermore, Wang et al [22] advocated for the development of open-ended tasks, such as clinical data analysis, and the establishment of case libraries showcasing AIGC applications (including examples of both accurate and inaccurate uses) within medical education, particularly in programming courses, to cultivate students’ critical utilization skills. Collectively, these studies and recommendations underscore the potential of customized AIGC tools in medical education to enhance critical thinking abilities and accommodate diverse student needs through personalized and varied learning approaches. However, the effectiveness of these tools is contingent upon their design and positioning to avert potential dependency risks. In contrast, Gong et al [23] examined the clinical applicability of customized LLMs in gastroenterology and found that the generic GPT-4o, equipped with retrieval-augmented generation capabilities, performed at an expert level, significantly surpassed the performance of customized models.

Montagna et al [24] conducted a systematic comparison of the efficacy of 3 distinct types of clinical decision support systems: traditional clinical practice guidelines, online knowledge bases (such as UpToDate), and LLMs (ChatGPT) in the context of medical student case resolution. Their findings indicated that while ChatGPT provided the quickest responses, clinical practice guidelines demonstrated superior accuracy. Concurrently, Mondal [25] explored the potential influence of a novel model of resource integration on the critical thinking abilities of medical students. Their survey revealed that medical students frequently utilize AI, particularly LLMs, to rapidly construct knowledge frameworks, subsequently validating detailed information through authoritative databases, such as PubMed. This hybrid approach, termed “LLM (Large Language Model) + Search Engine + Authoritative Database,” not only facilitates the efficient integration of multisource information but also fosters the development of students’ “multiresource integration ability.” This capability is essential for addressing potential obsolescence or inaccuracies in AIGC-generated information, while also empowering students to maintain critical thinking when confronted with diverse information sources, thereby enhancing their skills in identifying and verifying the accuracy and reliability of information. This innovative model offers medical students a more holistic educational pathway, ultimately aiding in the enhancement of their critical thinking and decision-making competencies within complex informational contexts.

Laupichler et al [26] conducted a comparative analysis of the quality of questions generated by ChatGPT-3.5 versus those created artificially, revealing that AI-generated questions were more susceptible to being “guessed,” while artificially constructed questions were more effective in accurately assessing students’ true capabilities. Thomae et al [27] further illustrated the benefits of engaging students in the design of solutions and subsequent comparisons with AI-generated alternatives within the same academic year, particularly in the context of a “placebo control condition” unit. This comparative exercise not only enables students to visually discern the strengths and weaknesses of AI but also promotes critical thinking through facilitated discussions. Additionally, Mondal [28] emphasized the necessity for students to actively articulate their evaluations of AI-generated information and engage in discourse regarding its potential inaccuracies and biases. This practice is not only fundamental to the development of critical thinking skills but also essential for navigating the limitations of AI technology. Collectively, these studies underscore that through the processes of comparison, evaluation, and discussion of AI-generated content, medical students can enhance their critical thinking abilities, thereby enabling them to apply AI technology more judiciously in their future medical practices.

In summary, AIGC functions simultaneously as an enhancer and a potential inhibitor of critical thinking among medical students. When utilized appropriately, AIGC can evolve into an effective “practice partner.” However, improper use may lead to adverse outcomes, such as AIGC dependency syndrome, cognitive stagnation, and a deterioration of academic integrity. Addressing these challenges requires an integrated approach comprising standardized AIGC patient simulations, personalized instructional guidance, multifaceted evaluation methods, and proactive assessment strategies.