Is your curriculum GenAI-proof? A method for GenAI impact assessment and a case study.

Is your curriculum GenAI-proof? A method for GenAI impact assessment and a case study - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Search Log in Primary site navigation Search Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice MedEdPublish (2016) . 2026 Apr 28;15:11. Originally published 2025 Mar 26. [Version 3] doi: 10.12688/mep.20815.3 Other versions Is your curriculum GenAI-proof? A method for GenAI impact assessment and a case study Remco Jongkind 1, a , Erik Elings 1 , Erik Joukes 2 , Tom Broens 1 , Hemmik Leopold 2 , Floris Wiesman 2 , Jennita Meinema 1 Author information Article notes Copyright and License information 1 Teaching and Learning Centre, Amsterdam UMC Location AMC, Amsterdam, North Holland, 1105AZ, The Netherlands 2 Medical Informatics, Amsterdam UMC Location AMC, Amsterdam, North Holland, 1105AZ, The Netherlands a Email: r.c.jongkind@amsterdamumc.nl No competing interests were disclosed. Roles Remco Jongkind : Conceptualization, Data Curation, Investigation, Methodology, Project Administration, Visualization, Writing – Original Draft Preparation, Writing – Review & Editing Erik Elings : Conceptualization, Investigation, Methodology, Project Administration, Writing – Review & Editing Erik Joukes : Conceptualization, Investigation, Project Administration, Writing – Review & Editing Tom Broens : Writing – Review & Editing Hemmik Leopold : Conceptualization, Investigation, Project Administration, Writing – Review & Editing Floris Wiesman : Conceptualization, Investigation, Project Administration, Writing – Review & Editing Jennita Meinema : Conceptualization, Investigation, Methodology, Project Administration, Writing – Original Draft Preparation, Writing – Review & Editing Accepted 2026 Apr 21; Collection date 2025. Copyright: © 2026 Jongkind R et al. This is an open access article distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. PMC Copyright notice PMCID: PMC12976610  PMID: 41821778 Version Changes Revised. Amendments from Version 2 In response to peer review comments, the following changes are made. Introduction. Two sentences have been added to strengthen the evidence base. The first, in the paragraph on universities' concerns about GenAI, cites Kasneci et al. (2023) and Perkins (2023) to establish that GenAI's capacity to complete academic tasks challenges traditional assessment validity and justifies a systematic curriculum-level response. The second, following the passage on students potentially not achieving learning outcomes, cites Kasneci et al. (2023) on the risk that repeated GenAI reliance may impair students' development of critical reasoning skills. Also in the Introduction, the unsupported hypothesis that GenAI would perform worse at lower Bloom's taxonomy levels has been removed. This assumption lacked citation and is contradicted by published evidence, including Nguyen Thanh et al. (2023), which found GenAI performs well at lower Bloom's levels, with "create" being its weakest area. The opening sentence on expected impact on writing and programming assessments has been retained. Methods. The description of Step 5 (Consolidation) has been substantially expanded to match the detail of Steps 1–4. The revised text describes the step's purpose, its integration into the annual PDCA cycle, the concrete mechanism (standardised GenAI questions in yearly exam evaluation reports followed by action plan meetings between course coordinators and educational management), and the relevant actors, inputs, and outputs. Abstract and Conclusion. "A potential blueprint" has been replaced with "a replicable framework" in both locations, better reflecting the adaptable nature of the method. Results. Sentence has been added after Figure 2 directing readers to the Open Science Framework repository (Jongkind et al., 2025; doi: 10.17605/OSF.IO/2YR6A) for the complete dataset of assessments and learning outcomes. As each assessment links to multiple learning outcomes, an enumerated list in the article would be unwieldy; readers seeking this detail are directed to the OSF dataset. Abstract Background Generative AI (GenAI) can complete tasks that previously could only be performed successfully by humans. Although GenAI provides many opportunities, it also poses risks such as invalid assessments and irrelevant learning outcomes. This article presents a broadly applicable method to (1) determine current assessment validity, (2) assess which learning outcomes are impacted by student GenAI use and (3) decide whether to alter assessment formats and/or learning outcomes. This is exemplified by a case-study on our Medical Informatics curriculum. We developed a five-step method to evaluate and address the impact of GenAI. Together with the respective lecturers, all assessments in the courses comprising a curriculum are analysed and subsequently adapted to address the impact of GenAI usage. Results 57% (55/96) of assessments, especially for courses involving writing and programming, were at risk of reduced validity and relevance. GenAI impact on assessment validity was closer related to the content and structure of assessments than to their complexity according to Bloom’s taxonomy. During educational retreats, lecturers discussed the relevance of impacted learning outcomes and whether students should be able to achieve them with or without GenAI. Furthermore, the results led to a plan to increase GenAI literacy and the intentional use by students during their education. For this purpose, course coordinators were asked to either adjust their manner of assessment to preclude GenAI use, or to alter learning outcomes to include GenAI use and literacy. For 64% of the impacted assignments (35/55) the assessment format was adapted and for 36% (20/55) the learning outcomes were adapted. Conclusion The majority of assessments in our curriculum were at risk of reduced assessment validity and relevance of learning outcomes, necessitating adaptations of either the manner of the assessment or learning outcomes. The proposed method and case-study offer a replicable framework for educational institutions facing similar challenges. Keywords: Generative AI, GenAI, Large Language Models, Curriculum adaptation, ChatGPT, assessment validity, AI literacy, learning outcome relevance. List of abbreviations GenAI Generative Artificial Intelligence LLM Large Language Models MI Medical Informatics UvA University of Amsterdam NVMO Nederlandse Vereniging voor Medisch Onderwijs - Dutch Society for Medical Education GPT Generative Pre-trained Transformer (specific to ChatGPT) Introduction In recent years, the application of artificial intelligence (AI) has shown great potential to perform certain tasks faster or better than human professionals. Initially, AI was mainly used for the analysis of data, sound or image, therefore limiting its impact to certain professions and tasks. However, with the introduction of ChatGPT in November 2022, it became evident that generative artificial intelligence (GenAI) was rapidly evolving to also assist with or take over professions previously thought to be exclusively human, such as medical informaticians or doctors. GenAI, such as large language models (LLM), refers to AI systems that use input data, or prompts, to produce new content. This digital content most prominently includes text, images, and audio, although various other forms of content can be requested. The current article focuses primarily on LLMs, particularly the most used variant ChatGPT; a GenAI designed to generate coherent and contextually relevant text based on user prompts. There are numerous examples showing that GenAI can take over and aid human processes in the workplace. For instance, Ayers et al. (2023) reported that ChatGPT v3.5 answered patients’ questions in an online forum more empathetically and accurately than doctors. ChatGPT v4, an upgraded version of ChatGPT v3.5, can effectively design, write and debug software code ( Azaiz et al. , 2023 ; Ellis et al. , 2024 ; Górecki, 2024 ; Idrisov & Schlippe, 2024 ; Kabir et al. , 2024 ; Kwon et al. , 2023 ; Palkowski & Gruzewski, 2024 ; Popovici, 2023 ; Reason et al. , 2024 ; Rodriguez et al. , 2024 ; Song et al. , 2024 ; Su & McMillan, 2024 ; Yeo et al. , 2024 ). ChatGPT can also be applied for modelling in health economics ( Reason et al. , 2024 ) and for reviewing mobile health apps ( Giunti & Doherty, 2024 ). In the domain of university education, GenAI is poised to transform teaching and assessment practices, raising new questions about its potential and challenges. ChatGPT and other GenAI can be employed as a tool for many aspects of academic writing, such as for abstracts ( Cheng et al. , 2023 ; Hwang et al. , 2024 ), introductions ( Sikander et al. , 2023 ) medical reports ( Zhou, 2023 ), case reports ( Buholayka et al. , 2023 ), scientific reviews ( Kacena et al. , 2024 ; Nazzal et al. , 2024 ), cover letters ( Deveci et al. , 2023 ), patient educational material ( Mondal et al. , 2023 ), essays ( Herbold et al. , 2023 ), research articles ( Abuyaman, 2023 ; Huespe et al. , 2023 ; Lund & Ting, 2023 ; Saito & Tsukiyama, 2024 ; Semrl et al. , 2023 ) and peer reviews ( Saad et al. , 2024 ). Its performance in finding and using references, however, remains subpar as of now ( Lechien et al. , 2024 ; Mugaanyi et al. , 2024 ). Nonetheless, considering the rapidly evolving capabilities of GenAI, this feature might be improved in the near future, further qualifying it as a writing tool ( Dell’Acqua et al. , 2023 ). GenAI has demonstrated the ability to pass various types of exams across educational domains. For example it has passed graduate level biomedical, physiology and biochemistry exams ( Ghosh & Bir, 2023 ; Stribling et al. , 2024 ; Subramani et al. , 2023 ), clinical informatics ( Kumah-Crystal et al. , 2023 ), graduate Python programming ( Ellis et al. , 2024 ), and a wide range of entry and specialist medical exams in a variety of countries such as Germany, the United States and the United Kingdom ( Cuthbert & Simpson, 2023 ; Farhat et al. , 2024 ; Griewing et al. , 2023 ; Kufel et al. , 2024 ; Li et al. , 2024 ; Lum, 2023 ; Meyer et al. , 2024 ; Sarangi et al. , 2024 ; Stengel et al. , 2024 ). Previous examples show that GenAI has abilities that overlap with learning outcomes present in many curricula. Most particularly, skills in knowledge reproduction, scientific writing, programming, data analysis and communication seem to be impacted. This is also the case for our Medical Informatics bachelor curriculum, in which students learn to understand basic medical subjects, methodologies of medical-scientific research, software development and healthcare organization. This overlap allows students to use GenAI in their studies, as has also been reported at other universities ( De Winter et al. , 2023 ). An informal anonymous poll among ~35 Medical Informatics students revealed that 95% admitted using ChatGPT in some form for their assessments. Another informal poll among ~100 of our third-year medical students provided a similar percentage (90%). This could be explained by students tending to choose the path of least resistance ( Abbas et al. , 2024 ). Uninformed use of GenAI is risky since students might be unaware of potential biases and incorrect answers that GenAI can sometimes provide. Furthermore, by using GenAI for assessments students might not achieve the intended learning outcomes relevant for their discipline. More broadly, research has raised concerns that repeated reliance on GenAI for academic tasks may reduce students’ engagement in higher-order thinking and impair the development of critical reasoning skills ( Kasneci et al ., 2023 ). Despite its potential, universities remain cautious about integrating GenAI into their curricula, citing concerns over academic integrity, assessment validity, and privacy. Indeed, published research has highlighted that GenAI’s capability to complete academic writing and knowledge-based tasks raises fundamental questions about the continued validity of traditional assessment formats ( Kasneci et al ., 2023 ; Perkins, 2023 ), underscoring the need for systematic approaches to evaluate and address this impact at the curriculum level. Additionally, the need to establish GenAI infrastructures that are accessible free of cost remains a significant challenge. Nevertheless, numerous universities have gradually started exploring pilot programs for GenAI management and usage. For instance, the University of Amsterdam (UvA) recently released policy guidelines stating that students may use GenAI if explicitly permitted for the specific course. Similarly, other universities have introduced policies stressing students’ responsibility for ensuring transparency and accountability in their work. In mid-2023, some of the lecturers began to adjust their assessments and teaching methods to incorporate the new developments in GenAI. This, however, posed the risk of disrupting curriculum cohesion, as not all courses made an effort to adopt GenAI usage. For example, students might initially pass an assessment using GenAI in one course, only to face restrictions on its use in a later course, creating discrepancies in the skills needed for them to complete the assessment. Therefore, we recognized the importance of evaluating the impact of GenAI on assessments and learning outcomes across entire study curricula. Our ultimate goal was to make any necessary adjustments to educational practices to ensure that assessments remain valid, learning outcomes stay relevant to future professional demands, and a consistent approach to GenAI use and policy is implemented. Ultimately, our goal is to safeguard the quality of the student learning experiences. Before the start of this study (May 2023), only initial reporting on the impact of GenAI on the assessment or course level was available ( Chaudhry et al. , 2023 ). However, a comprehensive and consistent approach to the evaluation of its impact across an entire curriculum was lacking. At the time, the only curriculum-wide initiative present was a methodology being developed by van Ooijen-van der Linden et al. (2023) , created to evaluate the impact of all forms of AI, not just GenAI. However, this methodology was still at an early stage of development, rendering it not yet usable for broader implementation. Additionally, it relied on time-consuming series of group discussions among lecturers, staff, and students, which presumably would not garner sufficient support from our teaching faculty. Informal inquiries within the Dutch Society for Medical Education (NVMO) working group on academic development and the Teaching and Learning Centres at other UvA faculties also failed to yield suitable methodologies. Consequently, we decided to develop our own method to assess the curriculum wide impact of GenAI on assessment validity and learning outcome relevance. In March 2023, it was hypothesized that assessments involving writing of non-academic products outside the classroom would be most impacted, followed by academic products and software coding. This article outlines the method and the results of the GenAI impact scan conducted on the Medical Informatics bachelor curriculum as a case study. It also details the steps we have taken to proof the curriculum for future developments, along with practical recommendations for other educational programs based on our experiences. GenAI impact scan method – determining the impact on assessment validity and relevance of learning outcomes In March 2023, five months after the launch of ChatGPT3.5, a five-step method was developed and subsequently implemented to evaluate the impact of GenAI at a curriculum level. The curriculum was adapted accordingly to preserve the constructive alignment of the courses. This method aimed to identify which types of assessments and learning objectives are most impacted by student use of GenAI. To address this issue, a multi-step approach was established ( Table 1 ). First, the urgency for action was determined (Step 1) to estimate the extent of the impact and create awareness and willingness for change among staff. Next, assessments where GenAI use was deemed impossible, such as on-site and practical examinations, were excluded (Step 2). This process incrementally funnels the learning assessments on their risk of being invalidated by GenAI and reduces the number of assessments up for evaluation, and hence the workload in later steps. Afterward, lecturers were consulted as experts on the subject matter to evaluate potential impact on their course (Step 3). To ensure coherent curriculum adjustments, a collaborative discussion with the lecturers was conducted (Step 4). Finally, given the rapid evolution of GenAI capabilities ( Dell’Acqua et al. , 2023 ), Steps 2, 3 and 4 are planned to repeat annually for each learning assessment. Table 1. Table describing the five steps of the GenAI impact scan method with the goal, the input needed for the step, the outcomes of the step and the involved actors. GenAI: Generative AI, PDCA: Plan-Do-Check-Act. Goal Input Outcomes Actors Step 1 Estimating the extent to which learning activities are impacted to determine urgency Set of learning activities used in the curriculum, such as workplace learning, projects, essays, etc. Set of used learning activities that are impacted by GenAI Educational advisors Step 2 Determining whether students have the opportunity to make the assessment with GenAI. This reduces the number of assessments that lecturers need to screen. Set of assessments in the curriculum Score that expresses the possibility of GenAI abuse per assessment. (No risk/Partial risk /High risk). Educational advisors Step 3 Determining to which degree GenAI can complete the assessments. This is to decide which assessments or learning outcomes to alter. Set of (partially) at risk assessments in the curriculum Score that expresses the degree of ability of GenAI to successfully complete the assessment per assessment (partially) at risk. Lecturers, educational advisors Step 4 Determining if and how to adapt the assessments or learning outcomes Set of assessments risk scores per assessment, set of affected learning outcomes of all courses. Exit competences were not considered due to their general nature. Decision per learning goal whether to keep, adapt or remove it. List of new learning outcomes that we wish to add to the curriculum, for example AI literacy. Lecturers, educational advisors, educational management Step 5 Ensuring the ongoing validity and relevance of assessments and learning outcomes by repeating step 2 and 3 yearly in the PDCA cycle. The assessments per course For each assessment an indication whether it is at risk and a score that expresses the degree of ability of GenAI to successfully complete the assessment. Lecturers Open in a new tab Step 1: Pre-scan In the first step, common educational formats are evaluated at a curriculum level and estimations are made on the degree to which GenAI can be used by students to complete these assessments. This step functions to give an initial impression of the impact of GenAI and to determine the urgency of the GenAI impact scan. Examples of such educational formats include team-based learning, project-based education, and workplace learning. We expect project- and case-based reports, programming take-home exams and academic writing to be aspects most impacted, since these assessments often occur without direct supervision and involve actions that GenAI is proficient in. Step 2: Quick scan For the second step, educational advisors identify assessments at risk of students using GenAI undetected to complete summative tasks, thereby failing to achieve the intended learning outcomes. Educational advisors evaluate all summative assessments within the curriculum, answering the following 3 questions for each assessment: • 1. Is the assessment conducted under supervision? When an assessment takes place under supervision, the likelihood of students displaying academic dishonesty is much lower ( Janke et al. , 2021 ; Yazici et al. , 2011 ). • 2. Does the course require intermediate products to be submitted before the final assessment? Intermediate products have been employed by many leading universities such as Caltech University, Cornell University and University of Tokyo to improve the authenticity and validity of the eventual summative assessments ( Moorhouse et al. , 2023 ). • 3. Is the weight of the assessment <10% of the final grade of the course? This question was mainly asked to prioritize which assessments to analyse and if necessary, to adapt first. If the answers to all three above questions are no, then the assessment is classified as high-risk. If the answer to question 1 is no, then the assessment is classified as partial risk. Otherwise, the assessment is classified as no risk. For all high or partial risk assessments a detailed scan is performed with priority being given to the high-risk assessments. Step 3: Detailed scan In the third step lecturers responsible for the assessments conclude whether the assessments that are deemed at risk can in fact be made with GenAI. In this step, two hands-on workshops are organized for the course coordinators and lecturers in which basic knowledge of GenAI and effective prompt writing is taught. For this workshop, the coordinators and lecturers should be provided access to a GenAI tool, which ideally mirrors one that students are likely to use. Additionally, it should preferably use an advanced frontier large language model, should not train on input data and should have functionalities such as uploading files, creating images, using plugins, and creating custom chatbots such as GPTs in ChatGPT. In these workshops, lecturers attempt to complete their own at-risk learning assessments with the help of GenAI. If they do not manage to complete all these assessments during the workshops they are asked to continue later and report to educational advisors via a central form within a predefined period of time. The lecturers are asked to classify their assessments on a Likert scale of 1 to 5 with levels as presented below. These labels were created by the educational advisors and tested for suitability on one first-year bachelor’s course and one first-year master’s course in medical informatics. These courses were chosen as representative for the curriculum by the educational management. Based on the initial tests, minor adjustments were made to the levels. Rating scale: 1. GenAI provides a perfect or very good output that can be used for the assessment. The prompt needed for GenAI to produce this output (almost) literally matches the assessment description. 2. GenAI provides a perfect or very good output that can be used for the assessment. The prompt needed for GenAI to produce this output has to be refined based on the assessment description. This requires knowledge of how to prompt effectively. 3. GenAI provides an insufficient solution based on solely the assessment description and effective prompts. Additional subject matter knowledge or skills are necessary for a perfect or good assessment product. 4. GenAI provides an insufficient output for the assessment, even with effective prompting and additional subject matter knowledge or skills. GenAI can only be used as inspiration; for example to brainstorm. 5. It is not possible to use GenAI for this assessment; for example, because actions must be taken in certain software systems which GenAI cannot access. The rating scale provides a measure of resistance a student may experience when using GenAI to help with an assessment, as students tend to choose the path of least resistance ( Abbas et al. , 2024 ). For example, for an assessment where (almost) literal copying of the assessment description results in a good outcome (level 1), the threshold for using GenAI is lower compared to an assessment that requires modifications to the prompt to achieve a good result (level 2). Therefore, we expect that each level on the scale represents a (non-linear) increase in usability and probability for a student to utilize GenAI for the assessment. Furthermore, the score indicates which skills or knowledge the students still need to complete the assessment, namely GenAI literacy skills or subject matter expertise. After rating all learning assessments in the curriculum on their level of AI usability, adjustments can be prioritized. The educational management decided that scores 1, 2 and 3 for assessments were undesirable and that either the manner of assessment or the learning outcomes should be adapted. Step 4: Curriculum adaptation Assessments with a score of 1, 2, and 3 in step 3 and their associated learning outcomes are categorized in clusters based on type of skill assessed. Cluster examples include: 1. Academic skills 2. Programming 3. Writing skills (academic and non-academic) 4. Meta-cognitive skills 5. Data (synthesizing, analysis, etc.) Subsequently, the impact of GenAI on the learning outcomes is discussed with all lecturers. The lecturers are divided into groups according to their expertise and asked to determine for each of their assigned learning outcomes whether they would like to start, adapt or continue teaching certain learning outcomes considering the advances in GenAI. In this context, “start” implies the addition of new learning goals to the curriculum (e.g. AI literacy), “adapt” implies adding GenAI use to the learning outcome and “continue”implies continuing to teach the skill without any adaptation for GenAI advances. For the category “adapt” the lecturers should indicate from which year of the study GenAI use should be allowed for that specific learning outcome. This ensures that the students learn basic knowledge and skill without GenAI, enabling them to critically evaluate the output when later allowed to use GenAI. This method is an adjusted format of the Start, Stop, Continue methodology from Cowan & George (1999) . Based on the output gathered from this meeting, the educational management formulates a policy for the coherent adjustment and addition of learning outcomes to the curriculum. Subsequently, course coordinators can then refer to this policy to adapt their at-risk assessments and learning outcomes. These adaptations fall into one of two categories: • Change the assessment format to prevent GenAI usage, ensuring students achieve the learning outcomes independently. • Adjust the learning outcomes to better reflect the (desired or future) situation in the workplace. This requires the learning outcomes, assessment and teaching to be adapted so that the students can achieve the new or adapted learning outcomes: for example, by learning how to make a business plan for an e-health app in co-creation with GenAI. Step 5: Consolidation Since the abilities of GenAI are rapidly improving ( Dell’Acqua et al. , 2023 ), it is important to keep evaluating the impact of GenAI use by students on assessment validity and learning outcome relevance. In practice, consolidation involves embedding a standardised GenAI review into the existing annual course evaluation cycle. Concretely, course coordinators are asked each year, as part of the regular exam evaluation report, whether their current assessments can still be completed using GenAI and, if so, what changes they propose to address this. These responses form the basis of a dedicated “action plan” meeting between the course coordinator and educational management held prior to the start of each academic year. The consolidation step thus formalises the ongoing nature of the GenAI impact scan: rather than being a one-time intervention, the method is designed to be self-renewing. The actors involved are the course coordinators, the educational advisors who coordinate the scan, and the educational management who oversee curriculum coherence and policy alignment. The input required is the set of assessments from the preceding academic year together with any new GenAI capabilities that have emerged; the output is an updated set of adapted or reaffirmed assessments and learning outcomes for the coming year. The process and decisions to be made during the five steps as described above are visualized in the process flow chart in Figure 1 . Figure 1. Process flow chart, presenting the questions to be answered or decisions to be made for each step. Open in a new tab Results – a case study from the Bachelor of Medical Informatics In the following section the case study from our own Medical Informatics bachelor curriculum will be highlighted to give an indication of the possible impact of student GenAI usage on the validity of assessments and relevance of learning outcomes. The results from this case study are summarized in Figure 2 . The curriculum consists of a total of 16 courses, 169 learning outcomes and 96 assessments. A complete overview of all individual assessments and assigned GenAI impact scores, is publicly available in the associated Open Science Framework repository ( Jongkind et al ., 2025 ). Figure 2. Flowchart describing the five steps of the impact scan method with the results per step. Open in a new tab Step 1 Pre-scan With the pre-scan, the characteristic elements of the Medical Informatics curriculum that would likely be impacted by GenAI were identified. These elements were gathered from our curriculum database Act-e ( Broens et al. , 2017 ). This is a database outlining all learning goals, learning activities and assessments in the curriculum. Subsequently, it was estimated to which degree these activities, or their assessments, are impacted by GenAI based on our personal experiences with the capabilities of GenAI and earlier literature ( Dell’Acqua et al. , 2023 ; George et al. , 2023 ; Kalla et al. , 2023 ). Workplace learning, for example, such as in medical residency, is relatively unaffected by GenAI as it primarily involves hands-on activities and verbal interactions, with minimal reliance on written assessments by the student. Our bachelor uses 19 different assessment types. Given that 74 out of 96 assessments (77%) in the bachelor’s program were project reports, academic writing, or programming take-home exams—types considered more vulnerable to GenAI—the next steps of the AI impact scan were considered more urgent. Step 2 Quick scan After identifying the learning activities that may be impacted by GenAI during the pre-scan, it became a high priority to determine which specific assessments were at risk of being completed with GenAI. This led to the implementation of the quick scan, where to focus was on identifying assessments vulnerable to unnoticed use of GenAI by students. Educational advisors determined which assessments were at risk because students had the opportunity to complete summative assessments unnoticed using GenAI and therefore potentially not reaching the required learning outcomes. In sum, a total of 71 (74%) assessments were deemed (partially) at risk for our bachelor curriculum. This included 57/96 (59%) summative assessments at partial risk and 14/96 (15%) which were deemed high-risk. These high numbers aligned with expectations based on the pre-scan because of the high amount of project and case-based education in the curriculum. Most notably, the academic reports, business plans, advice reports and take-home exams for programming were scored at risk. Step 3 Detailed scan Having established which assessments were potentially at risk through the quick scan, a more in-depth analysis was necessary to understand the extent of this risk. In the detailed scan, lecturers were consulted to evaluate how and to what degree GenAI could be used to complete these learning assessments. For this purpose, lecturers were offered 2 workshops during which they were trained in the use of GenAI. Subsequently, lecturers were provided with ChatGPT Team licences, which were GDPR-compliant and no training was done on the input data. During and following the workshop the lecturers determined on the 1-5 scale which of the (partially) at risk learning assessments could in fact be completed with GenAI. We selected ChatGPT (version 4) over other GenAI tools such as Claude, Gemini, Mistral, and Llama, assuming most students would use it due to its strong brand recognition. Additionally, ChatGPT consistently ranked as the most capable tool on LLM benchmark boards, such as UC Berkeley’s LMSys Chatbot Arena, during the study period (April 2023–June 2024). The five-point rating criteria as discussed in the method section were applied. Fifty-five of the 71 at-risk assessments had a score between 1 and 3 and could therefore to a certain degree be completed by GenAI. Thus, 55 out of the total 96 assessments (57%) in our curriculum were at risk of reduced assessment validity or learning goal relevance. Over the entire curriculum, an average AI impact score of 3.1/5 was found. This varied between year 1 (2.8/5), year 2 (3.5/5) and year 3 (2.9/5) ( Figure 3 , left). The average scores of at-risk assessments per course are displayed in Figure 3 on the right. Following the plenary meetings, the results of these assessments were discussed one-on-one by the curriculum committee with the coordinating staff of all courses in the curriculum. Figure 3. Two graphs depicting the GenAI impact scores assigned by lecturers, indicating to which degree GenAI can be used to make the assessments, per year (left) or course (right). Open in a new tab With 1 being GenAI able to completely make the assessment and 5 that no GenAI usage is possible. For a detailed description of the interpretation of each score see step 3 in the method section. The courses where no impact score is indicated had no assessments where students have the opportunity to make the assessments with GenAI. Each dot in the left graph represents an assessment, each dot in the right graph represents an average of the assessments across a course. An assessment can have a non-integer score if some elements of the assessment fall in one score bracket and some in the other. Step 4 Curriculum adaptation With the detailed scan revealing which assessments could indeed be (partially) completed using GenAI, it became necessary to address this identified impact. In response, we undertook a process of adapting the curriculum to mitigate the impact of GenAI on learning outcomes, resulting in a series of policy changes, assessment adjustments and learning goal adaptations. The first step in this process was to cluster the learning outcomes into the following six categories: 1. Subject matter knowledge 2. Academic skills 3. Programming 4. Writing skills (academic and non-academic) 5. Meta-cognitive skills 6. Data (e.g., synthesizing, analysis) Subsequently, an educational afternoon retreat of 3 hours with all the medical informatics lecturers was organized. Groups were assembled according to field of expertise and asked to determine for each of their assigned learning outcomes whether they would like to start, adapt or continue teaching these learning outcomes to students considering the advances in GenAI. Reasons mentioned to allow GenAI use for learning goals (“adapt”) were preparing students for the future workplace and teaching students AI literacy. Reasons for not allowing students to use GenAI for learning goals (“continue”) were the inherent value of a process, for example writing to learn to structure thoughts, and acquiring sufficient knowledge and understanding of the material to be able to critically evaluate the output. Each group was chaired by an educational advisor or a programme director. Based on the input from this afternoon retreat, the curriculum committee defined a policy on the usage of GenAI by students throughout the Bachelor of Medical Informatics. The majority of the learning outcomes (~80%) were indicated by lecturers as important to continue teaching, as these covered basic knowledge and were necessary to assess the correctness of the output of GenAI. For the remaining ~20% of the learning outcomes, lecturers indicated that these would benefit from (partial) adaptations to include GenAI use. Several learning outcomes were added, mainly pertaining to GenAI literacy and skills. These learning outcomes should enable students to effectively and responsibly use GenAI regarding privacy, data security, environmental impact, academic integrity, biases, didactic considerations (e.g., mental complacency) and output accuracy. In line with the policy of the University of Amsterdam, educational management recommends that first-year coordinators should limit or avoid the use of GenAI to ensure students establish a solid foundation in medical informatics. During this time, students are also trained to critically evaluate GenAI output for reliability and accuracy. In the second and third years, coordinators are encouraged to explore AI use for the learning goals and its potential to enhance learning, while ensuring it does not fully replace traditional methods. Furthermore, it is important that lecturers including the use of GenAI for the development of a specific skill clearly communicate this use to later lecturers in order to maintain constructive alignment across the curriculum. As a general principle, students should be informed that the use of GenAI is not permitted unless its permission for use in assessments is explicitly stated in the course manual. The most common changes to learning assessments are regarding the assessment format, precluding the use of GenAI. Out of the 55 assessments with an assigned impact score of 1–3, for 35 assessments (64%) the format was changed by the course coordinators. These changes most often included examinations under supervision, changing a paper to a presentation or podcast, making the assessment process-focused and adding a number of (formatively) assessed intermediate products. For the remaining 20 out of 55 assessments (36%) with a score of 1–3, which mainly included assessments in programming and non-academic writing, learning outcomes were altered by the course coordinators. The most common adjustments to learning outcomes included adjusting assessment weightings in grading rubrics, incorporating AI literacy, raising the difficulty of learning outcomes, and rephrasing outcomes to integrate GenAI usage. Step 5 Consolidation Once the curriculum had been adapted to the GenAI impact, a process for continuous adaptation was required as GenAI capabilities are rapidly increasing ( Dell’Acqua et al. , 2023 ). This was established in the consolidation step, during which a long-term strategy was developed to guarantee that the curriculum remains aligned with evolving GenAI capabilities. To ensure continuous adjustment, the quick scan (step 2), detailed scan (step 3) and curriculum adaptation (step 4) are integrated into the Plan Do Check Act (PDCA) cycle for each course. In a yearly exam evaluation report, course coordinators are asked questions on whether students can complete their assessments using GenAI and if so, what the proposed changes are. The answers to these questions will then be discussed in a subsequent “action plan” meeting between the course coordinator and the educational management preceding the start of each course. As this last analysis will be implemented in the upcoming academic year, no data is available to illustrate this step as of now. Conclusion & discussion The growing use of GenAI by students has necessitated a revaluation and adjustment of assessments and learning outcomes of curricula. This article describes a generalizable method to determine the validity of assessments and impact of GenAI on learning outcomes. It also proposes a process to systematically decide whether to alter the assessment formats or the learning outcomes. This approach is illustrated through a case study on the medical informatics curriculum revealing that 57% (55/96) of the learning assessments could be completed to a certain degree using ChatGPT. The usability of ChatGPT did not decrease in year 3 compared to year 1 of the bachelor whereas the Bloom levels of the assessments do increase ( Figure 4 ) as students progress. The average GenAI impact score was 2.9 in year 1, 3.5 in year 2, and 2.8 in year 3. Bloom levels do not appear to determine the performance of a specific learning assessment on our GenAI impact scale; conversely, the content or manner of assessment formulation seems to be of importance. Furthermore, the assessments were made with ChatGPT v4, which provides considerably fewer hallucinations than ChatGPT v3.5 and has higher performance in general (e.g. Chelli et al. , 2024 ), thus potentially improving its usability at lower Bloom levels. Similarly, as of spring 2023, other studies have been published showcasing either no difference in the performance of ChatGPT across different Bloom levels or an improved performance at lower Bloom levels ( Ghosh & Bir, 2023 ; Herrmann-Werner et al. , 2024 ). Figure 4. Shows the distribution of the complexity of learning outcomes, categorized by Bloom's taxonomy, across the curriculum database Act-e for courses 1.1–3.4. Open in a new tab The first number represents the year of the course, and the second represents the order of the course within that year. The essential challenge within a curriculum is to decide which learning outcomes students should achieve without GenAI to ensure they acquire basic knowledge, and which learning outcomes should allow students to use GenAI. For this, it is important to discern various types of GenAI use which may be applicable to the different learning outcomes. For instance, more basal learning outcomes, such as the ability to produce grammatically accurate text in a scientific tone, can be achieved through GenAI without significant consequences to academic integrity. Such usage is comparable to more traditional manners of superficial text reviewing, like human copy editors or built-in spell checker extensions. As these functions primarily aid with the format, but not the content, the requirement for the student to demonstrate critical, original thinking remains. Employing GenAI tools to produce original content, conversely, poses a risk for adequate academic development and independent functioning in the future workplace. It is up to lecturers to decide to what extent this manner of usage is acceptable for their learning outcomes in step 4. The decision should also take into account the consistency of learning goals across the curriculum. Research shows that students with access to GenAI reach performance scores up to 48 to 127% higher than students without access to GenAI. However, when presented with a similar task at a later assessment without the ability to use GenAI, the students that used GenAI score 17% lower than students who did initially not have access to GenAI ( Bastani et al. , 2024 ). The higher performance of students using GenAI, illustrates that prohibiting the use of GenAI is not effective, since students can be expected to have to use GenAI in the workplace to be competitive ( Dell’Acqua et al. , 2023 ). It is however important that such GenAI access is consistent or increases incrementally over the course of a curriculum, to prevent the setback experienced by students who acquired a skill using GenAI but are no longer allowed to use it in following courses. When deciding to allow GenAI use for learning assessments, students need to be educated to effectively and responsibly use GenAI with regards to privacy, data security, environmental impact, academic integrity, biases, output accuracy and didactic considerations such as mental complacency. There are some important considerations and notes regarding the case study. First, since our main aim was to adapt the curriculum and ensure adequate learning of skills and knowledge by our students, several learning assessments were already adjusted before or during our study. Consequently, the results provided in the case study do not represent a completely “pre-GenAI” curriculum and the reported impact of GenAI on the validity and relevance of assessments will be an underestimation. Additionally, the process by which lecturers evaluated the capability of ChatGPT to complete their assessments is subjective and dependent on the lecturers’ GenAI proficiency. To mitigate this subjectiveness and skill dependency, we provided two workshops in which we trained the lecturers in GenAI use. We also encouraged the lecturers to collaborate during the workshops on the assessments and discuss their verdict with each other and the educational advisors. Since most lecturers are medical informaticians with high digital literacy, they were well-prepared to engage with GenAI. However, when applying this method to other curricula, more extensive GenAI training may be necessary. A potential improvement could involve having student assistants attempt the assessments with GenAI to better simulate students’ proficiency with both the subject matter and GenAI. Beyond the impact on the learning outcomes and the (future) work field, GenAI also offers great opportunities and considerations for lecturers and students during the study itself. For instance, several studies show that GenAI tools can assist students to develop certain skills, such as academic writing( Kurniati & Fithriani, 2022 ; Wang, 2020 ; Zhao, 2023 ). However, other studies demonstrate or predict negative effects such as the development of cognitive biases, lower grades due to assessors’ prejudices against the use of AI, the required extra attention for critical appraisal of sources, and cognitive complacency ( Abbas et al. , 2024 ; Choudhury & Shamszare, 2023 ; Liu et al. , 2022 ; Lund & Wang, 2023 ). Therefore, it is important that more research is done regarding the potential cognitive effects of GenAI use on learning, specifically on which guardrails would work best to provide optimal learning. GenAI applications that don’t instantly give away answers but provide the right amount of challenge to encourage students to think, interact with the material and allow for reflection could be a solution ( Bastani et al. , 2024 ). To ensure students achieve the intended learning outcomes without GenAI a shift is needed from traditional written assessments to either supervised or more interactive forms of assessment, such as presentations, in-person exams, or process-focused projects that are difficult for AI to replicate ( Bower et al. , 2024 ; Chaudhry et al. , 2023 ; Kaldaras et al. , 2024 ; Zirar, 2023 ). A challenge this solution poses is that such methods generally tend to be more time consuming for faculty and more resource intensive, requiring examination or lecture rooms. Moreover, it should be acknowledged that writing assignments in their traditional way contribute significantly to academic development by providing a distinctive educational value apart from solely functioning as an assessment method. This is illustrated by a substantial multi-institutional study by Anderson et al. (2015) , who found that effective academic writing practice gives shape to fundamental skills which allow for deeper learning and higher perceived gains in learning and development. Such skills include the evaluation of sources, synthesizing theories on a more complex level, perspective-taking and application of learned knowledge in other contexts, as well as more practical, personal and social skills of problem-solving and eloquence. Likewise, being able to critically revise written text is an important skill at risk when different assessment types are implemented. For instance, exam assessments might rely predominantly on memorization and often do not allow for critical revision and integrative discussion of topics across different perspectives. This removes the interactive and meaning-making aspect of effective writing practice, as proposed by the literature. Writing additionally motivates the student to demonstrate unique thinking skills, enhancing creativity and individuality ( Herrington, 1981 ). For above-mentioned reasons, some writing assessments should be maintained in their original format, instead adapting learning goals to integrate GenAI usage. In this case, it is important to account for the potential increase in difficulty and scope of learning assessments, as students using GenAI could be expected to produce more extensive output. To prevent inequity, it is essential to provide students with equal access to GenAI platforms, ensuring that those unable to afford GenAI subscriptions are not disadvantaged. In conclusion, our findings advocate for a systematic and adaptive approach to curriculum development in response to GenAI use and its impact on assessment validity and relevance. The method and experiences outlined in this paper offer a replicable framework for educational institutions facing similar challenges. Ethics and consent This study adheres to the ethical principles outlined in the Declaration of Helsinki and complies with the relevant institutional and national ethical guidelines governing educational research. Ethical approval Since we analyze assessments structure and content in a learning management system, this study did not require formal ethical approval by the Ethical Review Board of our local institution (Amsterdam University Medical Centre) or the Ethical Review Board (ERB) of the Dutch Association for Medical Education (Nederlandse Vereniging van Medisch Onderwijs, NVMO). According to the Amsterdam UMC institutional guidelines , research must be reviewed by the Medical Ethical Review Board (METC) if it qualifies as medical-scientific research under the Medical Research Act (Wet Medisch-wetenschappelijk onderzoek, WMO). Research is considered subject to WMO if there is any risk of physical or psychological harm to participants. In this study there are no participants, therefore does not have to be reviewed by the ERB of the Amsterdam UMC. If a study does not fall under the WMO but focusses on education, ethical approval by the NVMO ERB could be required. The NVMO ERB policy states that ethical approval is not required as this study only involved collecting data on the structure and content of educational assessments from a learning management system, with no potential physical or psychological impact to individuals. This assessment was based on the following criteria: • The data collected is on the structure and content of the assessments and learning goals. • The data were collected as part of an existing quality assurance procedure within regular educational activities. • No additional effort was required from lecturers beyond their routine duties. • No personal or identifiable data were collected. • There was no risk of harm or disadvantage to any individuals involved. Consent and participant considerations This study did not involve human participants. The lecturers were not research participants but rather were engaged in their professional role as part of the routine quality assurance process. They were tasked by educational management to assess the impact of potential Generative AI (GenAI) use on learning goals and assessments. This was a standard aspect of their work responsibilities and would have occurred regardless of the study. As no individual lecturer data were collected and no intervention was conducted, informed consent was not required. The model informed consent form from the NVMO requires to ask for consent if personal data is collected from the participants. The study did not involve surveys, interviews, or any additional data collection beyond existing curriculum materials (e.g., assignment descriptions from Act-e). All data used in this study were fully anonymized and could not be traced back to individuals. Given that the research was conducted within the framework of routine educational quality assurance and complied with institutional and national guidelines, this study was exempt from formal ethical review. Confidentiality and data protection All data were handled in accordance with the General Data Protection Regulation (GDPR) and the Amsterdam UMC data protection policies. Data storage complied with institutional regulations, ensuring that no individuals could be identified in any published results. Acknowledgements We would like to thank all lecturers of the medical informatics curriculum for their active role in examining and adjusting their education in response to the impact of GenAI. We also thank Roeland van der Rijst and Myrthe Heikens for their insightful feedback, which significantly improved the structure and clarity of this manuscript. Funding Statement The author(s) declared that no grants were involved in supporting this work. [version 3; peer review: 3 approved] Data availability Open Science Framework: Is your curriculum GenAI-proof? A method for GenAI impact assessment and a case study. 10.17605/OSF.IO/2YR6A . ( Jongkind et al. , 2025 ) The project contains the following underlying data: • MI2024 data AI impact scan (GraphPad Prism file, in which the graphs were created). • Medical Informatics 2024 – results GenAI impact scan (Excel file containing the same results as in the GraphPad Prism file to allow for more universal access to the data). The data is available under the terms of the Creative Commons Zero “No rights reserved” data waiver (CC0 1.0 Public domain dedication) ( http://creativecommons.org/publicdomain/zero/1.0/ ). Software availability In the article the below generative AI tools are discussed. These tools can be accessed via the provided links. Archived versions of the software are not available. GenAI tool Proprietary or open source Software available from: ChatGPT Proprietary (with free options) https://chatgpt.com/ Claude Proprietary (with free options) https://claude.ai/ Gemini Proprietary (with free options) https://gemini.google.com/ Mistral Open Source initiative (OSI) approved open license software is under Apache 2.0-license. https://mistral.ai/ Llama Restricted Open License: Meta Llama 3.2 Community License https://www.llama.com/ Open in a new tab References Abbas M, Jam FA, Khan TI: Is it harmful or helpful? Examining the causes and consequences of generative AI usage among university students. Int J Educ Technol High Educ. 2024;21(1):10. 10.1186/s41239-024-00444-7 [ DOI ] [ Google Scholar ] Abuyaman O: Strengths and weaknesses of ChatGPT models for scientific writing about medical Vitamin B12: mixed methods study. JMIR Form Res. 2023;7:e49459. 10.2196/49459 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Anderson P, Anson CM, Gonyea RM, et al. : The contributions of writing to learning and development: Results from a large-scale multi-institutional study. Res. Teach. Engl. 2015;50(2):199–235. [ Google Scholar ] Ayers JW, Poliak A, Dredze M, et al. : Comparing physician and Artificial Intelligence chatbot responses to patient questions posted to a public social media forum. JAMA Intern Med. 2023;183(6):589–596. 10.1001/jamainternmed.2023.1838 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Azaiz I, Deckarm O, Strickroth S: AI-enhanced auto-correction of programming exercises: how effective is GPT-3.5? International Journal Of Engineering Pedagogy. 2023;13(8):67–83. 10.3991/ijep.v13i8.45621 [ DOI ] [ Google Scholar ] Bastani H, Bastani O, Sungu A, et al. : Generative AI can harm learning. (SSRN Scholarly Paper 4895486),2024. 10.2139/ssrn.4895486 [ DOI ] Bower M, Torrington J, Lai JWM, et al. : How should we change teaching and assessment in response to increasingly powerful generative Artificial Intelligence? Outcomes of the ChatGPT teacher survey. Educ Inf Technol. 2024;29:15403–15439. 10.1007/s10639-023-12405-0 [ DOI ] [ Google Scholar ] Broens T, Wiesman F, Jaspers M: Towards a medical informatics curricula visualization and analysis tool. 2017. Buholayka M, Zouabi R, Tadinada A: The readiness of ChatGPT to write scientific case reports independently: a comparative evaluation between human and Artificial Intelligence. Cureus. 2023;15(5):e39386. 10.7759/cureus.39386 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Chaudhry IS, Sarwary SAM, El Refae GA, et al. : Time to revisit existing student’s performance evaluation approach in higher education sector in a new era of ChatGPT – a case study. Cogent Educ. 2023;10(1):2210461. 10.1080/2331186X.2023.2210461 [ DOI ] [ Google Scholar ] Chelli M, Descamps J, Lavoué V, et al. : Hallucination rates and reference accuracy of ChatGPT and bard for systematic reviews: comparative analysis. J Med Internet Res. 2024;26:e53164. 10.2196/53164 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Cheng SL, Tsai SJ, Bai YM, et al. : Comparisons of quality, correctness, and similarity between ChatGPT-Generated and human-written abstracts for basic research: cross-sectional study. J Med Internet Res. 2023;25(1):e51229. 10.2196/51229 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Choudhury A, Shamszare H: Investigating the impact of user trust on the adoption and use of ChatGPT: survey analysis. J Med Internet Res. 2023;25:e47184. 10.2196/47184 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Cowan J, George J: A handbook of techniques for formative evaluation: mapping the students’ learning experience. Routledge;1999. [ Google Scholar ] Cuthbert R, Simpson AI: Artificial Intelligence in orthopaedics: can chat generative pre-trained transformer (ChatGPT) pass section 1 of the fellowship of the Royal College of Surgeons (Trauma & Orthopaedics) examination? Postgrad Med J. 2023;99(1176):1110–1114. 10.1093/postmj/qgad053 [ DOI ] [ PubMed ] [ Google Scholar ] De Winter JCF, Dodou D, Stienen AHA: ChatGPT in education: empowering educators through methods for recognition and assessment. Informatics. 2023;10(4):87. 10.3390/informatics10040087 [ DOI ] [ Google Scholar ] Dell’Acqua F, McFowland E, III, Mollick ER, et al. : Navigating the jagged technological frontier: field experimental evidence of the effects of AI on knowledge worker productivity and quality. (SSRN Scholarly Paper 4573321),2023. 10.2139/ssrn.4573321 [ DOI ] Deveci CD, Baker JJ, Sikander B, et al. : A comparison of cover letters written by ChatGPT-4 or humans. Dan Med J. 2023;70(12):A06230412. 38018708. [ PubMed ] [ Google Scholar ] Ellis ME, Casey KM, Hill G: ChatGPT and Python programming homework. Decis Sci. 2024;22(2):74–87. 10.1111/dsji.12306 [ DOI ] [ Google Scholar ] Farhat F, Chaudhry BM, Nadeem M, et al. : Evaluating large language models for the national premedical exam in India: comparative analysis of GPT-3.5, GPT-4, and Bard. JMIR Med Educ. 2024;10:e51523. 10.2196/51523 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] George AS, George ASH, Martin ASG: A review of ChatGPT AI’s impact on several business sectors. Partners Universal International Innovation Journal (PUIIJ). 2023;01(01):9–23. 10.5281/zenodo.7644359 [ DOI ] [ Google Scholar ] Ghosh A, Bir A: Evaluating ChatGPT’s ability to solve higher-order questions on the competency-based medical education curriculum in medical biochemistry. Cureus. 2023;15(4):e37023. 10.7759/cureus.37023 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Giunti G, Doherty CP: Cocreating an automated mHealth apps systematic review process with generative AI: design science research approach. JMIR Med Educ. 2024;10:e48949. 10.2196/48949 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Górecki J: Pair programming with ChatGPT for sampling and estimation of copulas. Comput Stat. 2024;39:3231–3261. 10.1007/s00180-023-01437-2 [ DOI ] [ Google Scholar ] Griewing S, Gremke N, Wagner U, et al. : Challenging ChatGPT 3.5 in senology-an assessment of concordance with breast cancer tumor board decision making. J Pers Med. 2023;13(10):1502. 10.3390/jpm13101502 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Herbold S, Hautli-Janisz A, Heuer U, et al. : A large-scale comparison of human-written versus ChatGPT-generated essays. Sci Rep. 2023;13(1):18617. 10.1038/s41598-023-45644-910616290 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Herrington AJ: Writing to learn: Writing across the disciplines. Coll. Engl. 1981;43(4):379–387. 10.2307/377126 [ DOI ] [ Google Scholar ] Herrmann-Werner A, Festl-Wietek T, Holderried F, et al. : Assessing ChatGPT’s mastery of bloom’s taxonomy using psychosomatic medicine exam questions: mixed-methods study. J Med Internet Res. 2024;26:e52113. 10.2196/52113 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Huespe IA, Echeverri J, Khalid A, et al. : Clinical research with large language models generated writing-clinical research with AI-assisted Writing (CRAW) study. Crit Care Explor. 2023;5(10):e0975. 10.1097/CCE.0000000000000975 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Hwang T, Aggarwal N, Khan PZ, et al. : Can ChatGPT assist authors with abstract writing in medical journals? Evaluating the quality of scientific abstracts generated by ChatGPT and original abstracts. PLoS One. 2024;19(2):e0297701. 10.1371/journal.pone.0297701 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Idrisov B, Schlippe T: Program code generation with generative AIs. Algorithms. 2024;17(2):62. 10.3390/a17020062 [ DOI ] [ Google Scholar ] Janke S, Rudert SC, Petersen Ä, et al. : Cheating in the wake of COVID-19: how dangerous is ad-hoc online testing for academic integrity? Comput Educ Open. 2021;2:100055. 10.1016/j.caeo.2021.100055 [ DOI ] [ Google Scholar ] Jongkind R, Elings E, Joukes E, et al. : GenAI impact scan data.[Dataset]. Open Science Framework. 2025. 10.17605/OSF.IO/2YR6A [ DOI ] Kabir S, Udo-Imeh DN, Kou B, et al. : Is stack overflow obsolete? An empirical study of the characteristics of ChatGPT answers to stack overflow questions. arXiv. 2024;arXiv:2308.02312. 10.48550/arXiv.2308.02312 [ DOI ] [ Google Scholar ] Kacena MA, Plotkin LI, Fehrenbacher JC: The use of Artificial Intelligence in writing scientific review articles. Curr Osteoporos Rep. 2024;22(1):115–121. 10.1007/s11914-023-00852-0 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Kaldaras L, Akaeze HO, Reckase MD: Developing valid assessments in the era of generative Artificial Intelligence. Front Educ. 2024;9:1399377. 10.3389/feduc.2024.1399377 [ DOI ] [ Google Scholar ] Kalla D, Smith N, Samaah F, et al. : Study and analysis of Chat GPT and its impact on different fields of study. (SSRN Scholarly Paper 4402499),2023. Kasneci E, Sessler K, Küchemann S, et al. : ChatGPT for good? On opportunities and challenges of large language models for education. Learn Individ Differ. 2023;103:102274. 10.1016/j.lindif.2023.102274 [ DOI ] [ Google Scholar ] Kufel J, Bielowka M, Rojek M, et al. : Assessing ChatGPT’s performance in national nuclear medicine specialty examination: an evaluative analysis. Iran J Nucl Med. 2024;32(1):60–65. 10.22034/IRJNM.2023.129434.1580 [ DOI ] [ Google Scholar ] Kumah-Crystal Y, Mankowitz S, Embi P, et al. : ChatGPT and the clinical informatics board examination: the end of unproctored maintenance of certification? J Am Med Inform Assoc. 2023;30(9):1558–1560. 10.1093/jamia/ocad104 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Kurniati EY, Fithriani R: Post-Graduate students’ perceptions of quillbot utilization in English academic writing class. Journal of English Language Teaching and Linguistics. 2022;7(3) Article 3. 10.21462/jeltl.v7i3.852 [ DOI ] [ Google Scholar ] Kwon S, Lee S, Kim T, et al. : Exploring LLM-based automated repairing of ansible script in edge-cloud infrastructures. Journal of web engineering. 3rd International Workshop on Big Data-Driven Edge Cloud Services (BECS) 2023;22(6):889–912. 10.13052/jwe1540-9589.2263 [ DOI ] [ Google Scholar ] Lechien JR, Briganti G, Vaira LA: Accuracy of ChatGPT-3.5 and -4 in providing scientific references in otolaryngology-head and neck surgery. Eur Arch Otorhinolaryngol. 2024;281(4):2159–2165. 10.1007/s00405-023-08441-8 [ DOI ] [ PubMed ] [ Google Scholar ] Li DJ, Kao YC, Tsai SJ, et al. : Comparing the performance of ChatGPT GPT-4, Bard, and Llama-2 in the Taiwan psychiatric licensing examination and in differential diagnosis with multi-center psychiatrists. Psychiatry Clin Neurosci. 2024;78(6):347–352. 10.1111/pcn.13656 [ DOI ] [ PubMed ] [ Google Scholar ] Liu Y, Mittal A, Yang D, et al. : Will AI console me when i lose my pet? Understanding perceptions of AI-mediated email writing. Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems. 2022:1–13. 10.1145/3491102.3517731 [ DOI ] [ Google Scholar ] Lum ZC: Can Artificial Intelligence pass the American board of orthopaedic surgery examination? Orthopaedic residents versus ChatGPT. Clin Orthop Relat Res. 2023;481(8):1623–1630. 10.1097/CORR.0000000000002704 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Lund B, Ting W: Chatting about ChatGPT: how may AI and GPT impact academia and libraries? (SSRN Scholarly Paper 4333415),2023. 10.2139/ssrn.4333415 [ DOI ] Lund BD, Wang T: Chatting about ChatGPT: how may AI and GPT impact academia and libraries? Library Hi Tech News. 2023;40(3):26–29. 10.1108/LHTN-01-2023-0009 [ DOI ] [ Google Scholar ] Meyer A, Riese J, Streichert T: Comparison of the performance of GPT-3.5 and GPT-4 with that of medical students on the written German medical licensing examination: observational study. JMIR Med Educ. 2024;10:e50965. 10.2196/50965 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Mondal H, Mondal S, Podder I: Using ChatGPT for writing articles for patients’ education for dermatological diseases: a pilot study. Indian Dermatol Online J. 2023;14(4):482–486. 10.4103/idoj.idoj_72_23 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Moorhouse BL, Yeo MA, Wan Y: Generative AI tools and assessment: Guidelines of the world’s top-ranking universities. Comput Educ Open. 2023;5:100151. 10.1016/j.caeo.2023.100151 [ DOI ] [ Google Scholar ] Mugaanyi J, Cai L, Cheng S, et al. : Evaluation of Large Language Model performance and reliability for citations and references in scholarly writing: cross-disciplinary study. J Med Internet Res. 2024;26:e52935. 10.2196/52935 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Nazzal MK, Morris AJ, Parker RS, et al. : Using AI to write a review article examining the role of the nervous system on skeletal homeostasis and fracture healing. Curr Osteoporos Rep. 2024;22(1):217–221. 10.1007/s11914-023-00854-y [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Palkowski M, Gruzewski M: GPT-driven source-to-source transformation for generating compilable parallel CUDA code for Nussinov’s algorithm. Electronics. 2024;13(3):488. 10.3390/electronics13030488 [ DOI ] [ Google Scholar ] Perkins M: Academic integrity considerations of AI large language models in the post-pandemic era: ChatGPT and beyond. J Univ Teach Learn Pract. 2023;20(2). 10.53761/1.20.02.07 [ DOI ] [ Google Scholar ] Popovici MD: ChatGPT in the classroom. Exploring its potential and limitations in a functional programming course. Int J Hum Comput Interact. 2023;40(22):7743–7754. 10.1080/10447318.2023.2269006 [ DOI ] [ Google Scholar ] Reason T, Rawlinson W, Langham J, et al. : Artificial Intelligence to automate health economic modelling: a case study to evaluate the potential application of Large Language Models. Pharmacoecon Open. 2024;8(2):191–203. 10.1007/s41669-024-00477-8 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Rodriguez DV, Lawrence K, Gonzalez J, et al. : Leveraging generative AI tools to support the development of digital solutions in health care research: case study. JMIR Hum Factors. 2024;11:e52885. 10.2196/52885 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Saad A, Jenko N, Ariyaratne S, et al. : Exploring the potential of ChatGPT in the peer review process: an observational study. Diabetes Metab Syndr. 2024;18(2):102946. 10.1016/j.dsx.2024.102946 [ DOI ] [ PubMed ] [ Google Scholar ] Saito H, Tsukiyama T: Use of Artificial Intelligence in manuscript preparation-AI as a co-author. Int J Periodontics Restorative Dent. 2024;1–12:38198433. 10.11607/prd.7022 [ DOI ] [ PubMed ] [ Google Scholar ] Sarangi PK, Narayan RK, Mohakud S, et al. : Assessing the capability of ChatGPT, Google Bard, and Microsoft Bing in solving radiology case vignettes. Indian J Radiol Imaging. 2024;34(2):276–282. 10.1055/s-0043-1777746 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Semrl N, Feigl S, Taumberger N, et al. : AI language models in human reproduction research: exploring ChatGPT’s potential to assist academic writing. Hum Reprod. 2023;38(12):2281–2288. 10.1093/humrep/dead207 [ DOI ] [ PubMed ] [ Google Scholar ] Sikander B, Baker JJ, Deveci CD, et al. : ChatGPT-4 and human researchers are equal in writing scientific introduction sections: a blinded, randomized, non-inferiority controlled study. Cureus. 2023;15(11):e49019. 10.7759/cureus.49019 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Song Y, Xie X, Xu B: When debugging encounters Artificial Intelligence: state of the art and open challenges. Sci China Inf Sci. 2024;67(4):141101. 10.1007/s11432-022-3803-9 [ DOI ] [ Google Scholar ] Stengel FC, Stienen MN, Ivanov M, et al. : Can AI pass the written European board examination in neurological surgery? - Ethical and practical issues. Brain Spine. 2024;4:102765. 10.1016/j.bas.2024.102765 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Stribling D, Xia Y, Amer MK, et al. : The model student: GPT-4 performance on graduate biomedical science exams. Sci Rep. 2024;14(1):5670. 38453979 10.1038/s41598-024-55568-7 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Su CY, McMillan C: Distilled GPT for source code summarization. Autom Softw Eng. 2024;31(1):22. 10.1007/s10515-024-00421-4 [ DOI ] [ Google Scholar ] Subramani M, Jaleel I, Krishna Mohan S: Evaluating the performance of ChatGPT in medical physiology university examination of phase I MBBS. Adv Physiol Educ. 2023;47(2):270–271. 10.1152/advan.00036.2023 [ DOI ] [ PubMed ] [ Google Scholar ] Linden L, Kootstra JG, Heijningen ME, et al. : AI4Students: methodiekontwikkeling om hogeschoolbreed AI-ready studenten op te leiden: Onderwijs Research Dagen 2023 (ORD).2023. Wang Z: Computer-assisted EFL writing and evaluations based on Artificial Intelligence: a case from a college reading and writing course. Library Hi Tech. 2020;40(1):80–97. 10.1108/LHT-05-2020-0113 [ DOI ] [ Google Scholar ] Yazici A, Yazici S, Erdem MS: Faculty and student perceptions on college cheating: evidence from Turkey. Educ Stud. 2011;37(2):221–231. 10.1080/03055698.2010.506321 [ DOI ] [ Google Scholar ] Yeo S, Ma YS, Kim SC, et al. : Framework for evaluating code generation ability of Large Language Models. ETRI J. 2024;46(1):106–117. 10.4218/etrij.2023-0357 [ DOI ] [ Google Scholar ] Zhao X: Leveraging Artificial Intelligence (AI) technology for english writing: introducing wordtune as a digital writing assistant for EFL writers. RELC J. 2023;54(3):890–894. 10.1177/00336882221094089 [ DOI ] [ Google Scholar ] Zhou Z: Evaluation of ChatGPT’s capabilities in medical report generation. Cureus. 2023;15(4):e37589. 10.7759/cureus.37589 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Zirar A: Exploring the impact of language models, such as ChatGPT, on student learning and assessment. Rev Educ. 2023;11(3):e3433. 10.1002/rev3.3433 [ DOI ] [ Google Scholar ] MedEdPublish (2016). 2026 May 6. doi: 10.21956/mep.23136.r45800 Reviewer response for version 3 Suriyaarachchige Nishan Silva 1, 2 Author information Copyright and License information 1 CHI Health, Claremont, Australia 2 Fellowship Support Program, Royal Australian College of General Practitioners, East Melbourne, Victoria, Australia Competing interests: No competing interests were disclosed. Roles Suriyaarachchige Nishan Silva : Referee Copyright: © 2026 Silva SN This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. PMC Copyright notice Thank you for accepting and including the suggested amendments constructively. This has strengthened the robustness of the flow and the content. Particularly, the changes made at the introduction have created a more impactful opening to the paper. Is the rationale for developing the new method (or application) clearly explained? Partly Is the description of the method technically sound? No Are the conclusions about the method and its performance adequately supported by the findings presented in the article? Partly If any results are presented, are all the source data underlying the results available to ensure full reproducibility? Partly Are sufficient details provided to allow replication of the method development and its use by others? No Reviewer Expertise: Health Informatician, Health professionals educationist I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard. MedEdPublish (2016). 2026 May 4. doi: 10.21956/mep.23136.r45801 Reviewer response for version 3 Nilesh Kumar Mitra 1 Author information Copyright and License information 1 IMU University, Kuala Lumpur, Malaysia Competing interests: No competing interests were disclosed. Roles Nilesh Kumar Mitra : Referee Copyright: © 2026 Mitra NK This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. PMC Copyright notice In this article, "A method for GenAI impact assessment and a case study", the authors have developed a five-step method to evaluate and address the impact of GenAI on assessment validity. The authors have involved several course coordinators to carry out the project and concluded that majority of assessments conducted using GenAI in their curriculum were at risk of reduced validity and this was related to the nature of learning outcomes. Following the last review the authors have added additional information on published literature under introduction to bring clarity to the research question. The authors have also described the 5th step consolidation in greater detail. Is the rationale for developing the new method (or application) clearly explained? Partly Is the description of the method technically sound? Yes Are the conclusions about the method and its performance adequately supported by the findings presented in the article? Yes If any results are presented, are all the source data underlying the results available to ensure full reproducibility? Yes Are sufficient details provided to allow replication of the method development and its use by others? Partly Reviewer Expertise: Technology enhanced learning, AI, Small group Learning, Online assessment I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard. MedEdPublish (2016). 2026 Mar 10. doi: 10.21956/mep.23018.r45469 Reviewer response for version 2 Nilesh Kumar Mitra 1 Author information Copyright and License information 1 IMU University, Kuala Lumpur, Malaysia Competing interests: No competing interests were disclosed. Roles Nilesh Kumar Mitra : Referee Copyright: © 2026 Mitra NK This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. PMC Copyright notice The article "Is your curriculum GenAI-proof? A method for GenAI impact assessment and a case study"  has been submitted under new educational method. The authors have claimed that they have developed a five step method to evaluate and address the impact of Generative AI on assessment. From the literature review described in the article, it is evident that comparable literature of impact assessment is not significant in number. It is clear that the authors could not append enough evidence about GenAI's impact on assessment and learning outcomes. The academic literature should describe the necessity of a research or study based on evidence available in the published literature. The authors hypthothesized that GenAi would perform worse at the lower Bloom's taxonomy of remembering and understanding. There is no citation available following this hypothesis. Academics should not append assumptions in the scientific literature. In 2023 Australian Journal of Educational Technology published an article written by Nguyen Thanh et al. which established that GenAI performed very well at lower level of Bloom's taxonomy with "create" being weakest level of performance. Following this article there are various publications, which have shown GenAI performance in creation of assessment tool and support of students in formative assessment involving learning outcomes with lower level of Bloom's taxonomy. We  understand the concern of authors about students losing critical thinking ability and thereby not being able to complete the learning outcomes. There are ample literature available presently related to this aspect of GenAI assisted assessment. It was expected that the authors would cite suitable references to establish the necessity of this research. Methods: The authors highlighted five-step method to evaluate and address the impact of GenAI. These five steps are: 1.Pre-scan 2. Quick scan 3. Detailed scan 4. Curriculum adaptation 5. Conslodation Let us look at the terminology used in describing the steps and whether prospective reader would be able to undestand the framework with this terminology. Pre-scan step describes evaluation of educational formats before starting the scanning of assessments. Quick Scan step describes identification of at risk assessments where use of GenAI remains undetected. Detailed scan describes the trial by the lecturers to test whether the assessment can be completed effectively with the help of GenAI. Five rating scales have been developed to classify the effectiveness of GenAI to support the assessments. Curriculum adaptation step helps to categorize learning outcomes and related skills. In contrast to clear description of first four steps, the description of fifth step consolidation is sketchy. The authors have mentioned that the proposed method and case-study offer a potential blueprint for educational institutions to follow. A perfect blue print is transparent, best represented in a tabular manner with steps starting from objectives and ending with steps of implementation with descriptions of steps, allowing easy implemntation. Results A birds-eye view of results is clearly depicted using Figure 2 where no. of assessments involved in each step is described in a flow chart. However, a list of learning outcomes written in text will allow the prospective reader to understand about the nature of outcome and relationship with the use of GenAI and its effect. Although Figure 4 has described the outcomes progress as per taxonomy. Discussion and Conclusion Compared to Introduction, this chapter has been written with a lot of clarity with sufficient evidence of literature. The authors described the process of adoption of GenAI in trasnformation of teaching process. Impact: This article is an excellent fit for new educational method. With a few changes the article can be used as a guide for GenAI impact assessment. Is the rationale for developing the new method (or application) clearly explained? Partly Is the description of the method technically sound? Yes Are the conclusions about the method and its performance adequately supported by the findings presented in the article? Yes If any results are presented, are all the source data underlying the results available to ensure full reproducibility? Yes Are sufficient details provided to allow replication of the method development and its use by others? Partly Reviewer Expertise: Technology enhanced learning, AI, Small group Learning, Online assessment I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above. MedEdPublish (2016). 2025 Aug 14. doi: 10.21956/mep.22282.r42321 Reviewer response for version 1 Suriyaarachchige Nishan Silva 1, 2 Author information Copyright and License information 1 CHI Health, Claremont, Australia 2 Fellowship Support Program, Royal Australian College of General Practitioners, East Melbourne, Victoria, Australia Competing interests: No competing interests were disclosed. Roles Suriyaarachchige Nishan Silva : Referee Copyright: © 2025 Silva SN This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. PMC Copyright notice It would have been better to elaborate on the five step method to evaluate and address the impact of GenAI. How did you arrive at the 57% of assessments were at risk of reduced validity and relevance?  It is not clear the assessment format mentioned in the second para of results is the same that was adopted by 64%. How was this decided? Is the rationale for developing the new method (or application) clearly explained? Partly Is the description of the method technically sound? No Are the conclusions about the method and its performance adequately supported by the findings presented in the article? Partly If any results are presented, are all the source data underlying the results available to ensure full reproducibility? Partly Are sufficient details provided to allow replication of the method development and its use by others? No Reviewer Expertise: Health Informatician, Health professionals educationist I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above. MedEdPublish (2016). 2026 Jan 21. Remco Jongkind 1 Author information Copyright and License information 1 Teaching & Learning Centre, Amsterdam UMC Locatie AMC, Amsterdam, North Holland, The Netherlands Competing interests: No competing interests were disclosed. PMC Copyright notice Dear Suriyaarachchige, Thank you for your useful feedback. We elaborated more on the five step method to improve clarity in the method section. Regarding the 57%, we included a clarification on page 21 on how we came to the 57%. Namely 55 out of 96 assessments were at risk. Furthermore on page 22 we included reasons for the choices to change assessment format or learning goal. Lastly we improved the overall language quality of the manuscript for better flow and clarity. Among others to ensure the numerous percentages regarding assessments at risk and lecturer decisions are now more transparent and easier to interpret within the narrative. Best wishes, Remco MedEdPublish (2016). 2025 May 2. doi: 10.21956/mep.22282.r41411 Reviewer response for version 1 Vijay Rajput 1 , Mohammadali Shoja 2 Author information Copyright and License information 1 Nova Southeastern University, Fort Lauderdale, USA 2 Medical Education, Nova Southeastern University Dr Kiran C Patel College of Allopathic Medicine (Ringgold ID: 503842), Fort Lauderdale, Florida, USA Competing interests: No competing interests were disclosed. Roles Vijay Rajput : Referee Mohammadali Shoja : Co-referee Copyright: © 2025 Rajput V and Shoja M This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. PMC Copyright notice We read with great interest the manuscript by Jongkind and colleagues, which explores a timely and highly relevant issue: the impact of generative AI on the validity of student assessment and the objectivity of learning outcomes. The authors propose a structured strategy for evaluating how generative AI intersects with learning objectives and assessment design across a curriculum. This is an important and forward-thinking endeavor, particularly given the rapid integration of AI technologies in academic settings. That said, we found that the overall language quality of the manuscript could be significantly strengthened to better support its strong conceptual contributions. At present, the writing lacks fluency in places, and some word choices are imprecise, which at times makes it difficult to fully grasp the intended meaning. For example, we occasionally needed to reread sentences to discern their context, and in some cases, the phrasing led to unintended interpretations. One such instance appears in the section titled “Step 1: Pre-scan,” where the authors write, “… it is estimated to which degree generative AI can be used to make these assessments.” This wording might unintentionally suggest that educators are using generative AI to construct assessments, rather than assessing the extent to which students might use AI to complete them—presumably the intended meaning. Similar issues appear throughout the paper and affect the overall clarity and accessibility of the work. Nonetheless, the substance of the study is both thoughtful and commendable. The authors present a compelling and well-structured framework for tackling the pedagogical challenges posed by generative AI in education. Their suggestion that learning outcomes may need to be assessed through more interactive or supervised formats is reasonable and well-supported. However, it may be helpful to more explicitly acknowledge the distinctive educational value of traditional written assessments—particularly those completed over time—which can foster critical thinking, reflection, and synthesis in ways that are not always replicable in supervised environments. In addition, the manuscript might benefit from addressing the growing importance of generative AI literacy among students. As educational institutions adapt to this new landscape, it will be essential to provide clear guidance on when and how AI tools can be ethically and effectively used. If restrictions are placed on AI usage in certain contexts, what enforcement mechanisms are available? These practical considerations would enhance the applicability of the proposed framework. Lastly, it may be helpful to distinguish more clearly between different types of AI use. While employing generative AI to produce ideas or content in place of a student’s own work raises valid concerns about academic integrity, using AI tools to refine language, improve grammar, or enhance structure may be more analogous to the support offered by a writing center or copyeditor—particularly for students writing in a second language. A nuanced discussion of these differences would enrich the paper and help inform more equitable and thoughtful policy development. In summary, this manuscript takes on a pressing and important challenge in higher education and offers a promising framework for addressing it. With careful revision to improve the clarity of the language and a more detailed consideration of implementation and ethical nuances, the paper could make a valuable contribution to the evolving conversation around generative AI in education. Is the rationale for developing the new method (or application) clearly explained? Partly Is the description of the method technically sound? Yes Are the conclusions about the method and its performance adequately supported by the findings presented in the article? Yes If any results are presented, are all the source data underlying the results available to ensure full reproducibility? Yes Are sufficient details provided to allow replication of the method development and its use by others? Yes Reviewer Expertise: Medical Education, AI in education, ethics, professionalism We confirm that we have read this submission and believe that we have an appropriate level of expertise to confirm that it is of an acceptable scientific standard. MedEdPublish (2016). 2026 Jan 21. Remco Jongkind 1 Author information Copyright and License information 1 Teaching & Learning Centre, Amsterdam UMC Locatie AMC, Amsterdam, North Holland, The Netherlands Competing interests: No competing interests were disclosed. PMC Copyright notice Dear Vijay and Mohammadali, Thank you for your elaborate and useful feedback! Based on your mentioned points of improvement we have implemented several enhancements. Below we provide details per point of improvement mentioned. Overall language quality of the manuscript The overall language quality has been refined for better flow, improving also the sentence mentioned in step 1. Furthermore we ensured the numerous percentages regarding assessments at risk and lecturer decisions are now more transparent and easier to interpret within the narrative. Distinctive educational value of traditional written assessments We have more explicitly discussed the inherent educational value of traditional written assessments. Drawing on pedagogical literature, we discuss on page 29 that the process of writing serves as a critical "meaning-making" exercise that fosters deep learning and individual creativity—qualities that risk being bypassed if outsourced to GenAI. Importance of generative AI literacy among students . The manuscript now elaborates more on the importance of AI literacy when allowing GenAI use for learning goals. We explain, a.o. on page 23 that if GenAI is permitted, it must be accompanied by explicit instruction on its limitations and ethical use. Enforcement mechanisms If restrictions are placed on AI usage in certain contexts We have provided more detail on potential alternative assessment formats (e.g. on page 24). We describe specific strategies, such as in-class reflections, oral examinations, and process-oriented grading, designed to ensure that students cannot use GenAI to circumvent achieving learning goals. Distinguishing between types of GenAI use for learning goals On page 26-27 we have introduced a clearer distinction between using GenAI for linguistic refinement (e.g., grammar and flow) versus content generation (e.g., idea synthesis and creative output). Best wishes, Remco Jongkind Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Data Citations Jongkind R, Elings E, Joukes E, et al. : GenAI impact scan data.[Dataset]. Open Science Framework. 2025. 10.17605/OSF.IO/2YR6A [ DOI ] Data Availability Statement Open Science Framework: Is your curriculum GenAI-proof? A method for GenAI impact assessment and a case study. 10.17605/OSF.IO/2YR6A . ( Jongkind et al. , 2025 ) The project contains the following underlying data: • MI2024 data AI impact scan (GraphPad Prism file, in which the graphs were created). • Medical Informatics 2024 – results GenAI impact scan (Excel file containing the same results as in the GraphPad Prism file to allow for more universal access to the data). The data is available under the terms of the Creative Commons Zero “No rights reserved” data waiver (CC0 1.0 Public domain dedication) ( http://creativecommons.org/publicdomain/zero/1.0/ ). Articles from MedEdPublish are provided here courtesy of Association for Medical Education in Europe ACTIONS View on publisher site PDF (4.3 MB) Cite Collections Permalink RESOURCES Similar articles Cited by other articles Links to NCBI Databases Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• Ensuring best practice in genomics education and evaluation: reporting item standards for education and its evaluation in genomics (RISE2 Genomics)#329991
• Have the Agile Principles Endured? An Empirical Investigation Post 20th Anniversary of the Agile Manifesto (2001)#497613
• The case study approach#6696
• <i>PLoS Genetics</i> Issue Image | Vol. 22(9) October 2026#1167704
• Technical Compliance Ratings - 5th round (2022 Methodology) - Public version#443276