Faculty perceptions of ChatGPT on academic integrity and institutional roles in higher education Authors: Salem Alalwani Published: 2026 ABSTRACT AbstractThe rapid expansion of generative artificial intelligence (GenAI) tools such as ChatGPT has raised growing concerns in higher education, particularly regarding assessment practices and academic integrity. This study examines faculty perceptions of ChatGPT in public universities in Saudi Arabia, with a focus on differences related to academic rank and institutional role. Using a sequential explanatory mixed-methods design, survey data were collected from 202 faculty members across eight public universities and analyzed using descriptive statistics and the Kruskal–Wallis test. Qualitative responses were analyzed thematically to explain and contextualize the quantitative findings. The results indicate variation in faculty perceptions. Assistant and associate professors reported higher concern regarding issues such as cheating, unequal evaluation, and the impact of AI on student performance and skills. Qualitative findings help explain these patterns by highlighting concerns about students’ reliance on AI-generated content and its effects on writing and learning processes. Faculty members in administrative roles emphasized institutional challenges, particularly the need for clearer guidelines on AI use in assessment and academic integrity procedures. The findings suggest that faculty perceptions of generative AI are shaped by professional responsibilities within specific institutional contexts. Practical implications include the need to develop clear institutional guidelines for AI use, redesign assessment practices to reduce misuse, and provide targeted faculty training on AI-aware teaching and evaluation strategies.Keywords: ChatGPT, Higher education, Faculty perceptions, Academic integrity, Assessment practicesSubject terms: Education, Information systems and information technology, Language and linguistics, Language and linguistics Introduction Generative artificial intelligence tools can produce text comparable to student work. This raises fundamental questions about how learning is assessed in higher education. Early work focused on the technical capabilities of these systems, while more recent research has shifted toward their pedagogical and ethical implications, particularly in relation to how knowledge is produced, assessed, and verified1,2. Within this broader context, artificial intelligence is widely viewed as a transformative force in education, reshaping pedagogical practices, learning processes, and assessment approaches3. Recent studies highlight different dimensions of this shift. Giannakos et al.4 examine changes in teaching and learning processes in AI-supported environments. Other research emphasizes both the opportunities and challenges of ChatGPT in academia, including its potential to enhance learning alongside concerns related to academic integrity and assessment practices5. Shata and Hartley6 further draw attention to institutional uncertainty and the need for clearer guidance in integrating AI tools. In a similar vein, Corbin et al.7 and Cassidy8 highlight the implications of generative AI for assessment practices and the evolving responsibilities of faculty in higher education. Evidence points to both the benefits and risks associated with generative AI in higher education. These tools can support feedback, content development, and instructional efficiency, while also raising concerns related to academic integrity, authorship, and the reliability of assessment practices9. At the same time, AI has the potential to promote deeper learning and more meaningful engagement with academic content10. However, recent evidence suggests that advanced AI systems can generate outputs comparable to student work, making it increasingly difficult to determine whether submitted work reflects independent learning11. Recent research suggests that the impact of generative AI extends beyond teaching and assessment to students’ learning experiences. Evidence from recent systematic reviews and meta-analyses indicates that, when appropriately integrated into higher education, generative AI can enhance academic achievement, language skills, learning motivation, and higher order thinking skills while supporting more personalized and flexible learning experiences12,13. However, recent evidence highlights important educational and ethical challenges associated with students’ use of generative AI. Although ChatGPT can facilitate understanding of complex concepts, idea generation, academic writing, and learning efficiency, concerns remain regarding overreliance on AI-generated content, reduced critical thinking, plagiarism, and privacy. These challenges underscore the need for universities to establish clear institutional policies, strengthen students’ digital literacy, and promote the responsible and ethical use of generative AI in higher education4,14. In the Saudi Arabian context, interest in artificial intelligence in education has increased in recent years. Empirical evidence indicates that students demonstrate high levels of awareness and acceptance of AI tools, alongside recognition of their potential benefits and associated challenges15. At the same time, although AI tools are widely used, levels of readiness and formal preparation remain uneven, highlighting a gap between use and understanding16. At the institutional level, the integration of artificial intelligence into educational practice in Saudi universities remains uneven, particularly in terms of structured implementation and institutional coordination16,17. Prior evidence also shows that the use of AI in teacher preparation programs is constrained by factors such as limited resources, insufficient training, and the absence of clear strategic direction16. In the Saudi context, centralized governance structures and evolving institutional policies may shape how faculty interpret and regulate the use of generative AI in teaching and assessment. However, this study does not examine institutional structures directly. It focuses on how differences in faculty roles and responsibilities relate to variation in perceptions. Prior studies in Saudi Arabia have focused on awareness, attitudes, readiness, and patterns of AI use. Less attention has been given to how generative AI is understood within core academic processes, especially assessment and academic integrity. There is also limited understanding of how perceptions differ by academic role, experience, and responsibility, and how these differences shape practice. This study shifts the focus from general awareness to interpretation within academic work. It extends existing research by examining how generative AI is understood within assessment and academic integrity, and by identifying how these interpretations vary across faculty roles and professional experience. Addressing this gap is important because generative AI challenges established assumptions about assessment, authorship, and academic responsibility in higher education. In this study, faculty perceptions are conceptualized as multidimensional evaluations reflecting how academic staff interpret, assess, and respond to the use of generative AI in higher education. Rather than treating perceptions as a general attitudinal construct, the study focuses on two interrelated dimensions: academic integrity and institutional roles. Academic integrity captures concerns related to authorship, assessment validity, and the ethical use of AI-generated content, while institutional roles reflect expectations regarding policy, governance, and faculty responsibility in regulating AI use. These dimensions provide an organizing framework for examining how faculty perceptions vary across roles and professional contexts. Saudi universities are expanding and are under increasing pressure to enhance quality, accountability, and research performance. In this context, faculty members play a central role in interpreting and applying emerging technologies in ways that align with institutional expectations. Their responses to generative AI are therefore shaped by teaching responsibilities, assessment practices, and institutional roles. This study examines faculty perceptions of ChatGPT in Saudi universities using a sequential explanatory mixed-methods design. This approach was adopted to address the complexity of the research problem, particularly the need to move beyond descriptive accounts of faculty attitudes toward a more nuanced understanding of how these perceptions are formed within institutional contexts. The quantitative phase identifies general patterns and differences in faculty perceptions, while the qualitative phase provides deeper insight into how participants interpret and explain these patterns in relation to academic integrity and assessment practices. By integrating both strands, the study offers a more comprehensive account of faculty responses to generative AI, which cannot be captured through a single-method approach. The study is guided by the following research questions: RQ1: To what extent do faculty members perceive challenges related to the use of ChatGPT in Saudi universities, and how do they interpret them? RQ1: To what extent do faculty members perceive challenges related to the use of ChatGPT in Saudi universities, and how do they interpret them? RQ2: To what extent do faculty members perceive the effectiveness of solutions for addressing these challenges, and how do they explain them? RQ2: To what extent do faculty members perceive the effectiveness of solutions for addressing these challenges, and how do they explain them? RQ3: Are there statistically significant differences in faculty members’ perceptions of challenges and proposed solutions based on demographic variables such as academic rank, teaching experience, and institutional role? RQ3: Are there statistically significant differences in faculty members’ perceptions of challenges and proposed solutions based on demographic variables such as academic rank, teaching experience, and institutional role? Generative artificial intelligence (AI) refers to a class of systems capable of producing human-like content, including text, in response to user prompts (OECD, 2023)18. In higher education, these technologies are increasingly shaping teaching, learning, and assessment practices, while also raising important ethical and governance-related challenges. Within this broader category, ChatGPT—developed by OpenAI—is treated in this study as a representative large language model-based application, examined in relation to its use in academic contexts by both students and faculty19. Academic integrity refers to adherence to ethical standards in academic work, including honest authorship and the responsible use of digital tools. In AI-mediated contexts, these standards are increasingly challenged, particularly as tools such as ChatGPT can generate human-like outputs that may be presented as students’ own work20. This raises concerns about AI-assisted plagiarism and academic misconduct in higher education9, as well as broader issues related to authorship and transparency in digital learning environments21. Assessment validity refers to the extent to which assessment practices accurately reflect students’ knowledge and performance. This becomes more complex when AI-generated responses obscure students’ independent understanding22, making it difficult to determine whether submitted work represents authentic learning. Such challenges have been highlighted in relation to ChatGPT, particularly regarding the credibility and fairness of evaluation processes9, as well as ongoing shifts in assessment practices (Korseberg & Elken, 2024)26. Institutional roles refer to the responsibilities of universities in developing policies and assessment practices that regulate the use of generative AI and maintain the integrity and credibility of academic processes23. Together, these concepts provide a basis for examining how faculty interpret the use of generative AI within teaching, assessment, and institutional contexts. Literature review Faculty perceptions of generative AI in higher education Recent scholarship on generative artificial intelligence (GenAI) in higher education has moved beyond a focus on the technical capabilities of AI tools toward their pedagogical, ethical, and institutional implications1,2,24. Within this literature, faculty perceptions play an important role in shaping how generative AI is interpreted, adopted, and regulated in academic contexts25,26. This reflects a broader shift in understanding AI integration as a socially and professionally situated process. Empirical studies across higher education contexts show that faculty attitudes toward generative artificial intelligence (GenAI) are often ambivalent. Academic staff recognize potential benefits, including improved feedback and instructional efficiency27, as well as support for personalized learning4. At the same time, they report concerns related to academic misconduct and reduced critical thinking6, in addition to increased workload and uncertainty about appropriate use. Research has also examined learners’ attitudes toward ChatGPT-assisted language learning, highlighting factors that shape student engagement and acceptance of AI tools in higher education28. Some studies further suggest that these perceptions may vary across academic rank and institutional roles29, with emerging evidence also indicating differences linked to institutional context and professional responsibilities30. However, the available evidence remains limited, particularly in the Saudi higher education context, where relatively little is known about how such perceptions are connected to academic responsibilities and assessment practices31. Generative AI offers clear gains in efficiency and feedback, but it also raises concerns about academic integrity, assessment validity, and skill development. Faculty perceptions appear to be shaped by their involvement in teaching and assessment rather than reflecting a single, shared position. This study examines how these perceptions vary across academic rank, years of experience, and administrative position within the Saudi higher education context. Academic integrity, assessment, and pedagogical concerns Academic integrity represents a central concern in relation to the use of generative artificial intelligence in higher education. Early studies identified concerns that tools such as ChatGPT may enable plagiarism20, contract cheating32, and unauthorized assistance in assessments7. These developments raise questions about the reliability of traditional assessment practices and the extent to which submitted work reflects students’ independent learning33,34 These concerns are compounded by the limitations of AI-detection technologies. Existing research highlights skepticism regarding their accuracy35 and fairness36, as well as uncertainty about their ability to distinguish between original and AI-generated content. More recent research has also highlighted broader limitations associated with AI systems in educational contexts, including concerns related to reliability, instructional effectiveness, and the challenges of integrating AI into complex learning environments, which continue to shape faculty perceptions and patterns of AI use37. This uncertainty has direct implications for teaching and learning. When the authenticity of student work cannot be reliably verified, attention may shift from supporting learning processes to managing potential misuse. In this context, pedagogical risks emerge. Over-reliance on AI-generated content may reduce students’ ability to think critically27, write independently, and solve problems38. These risks remain closely linked to ongoing challenges in maintaining academic integrity under conditions of limited detection reliability. Collectively, these studies indicate that concerns about academic integrity, the limitations of detection technologies, and pedagogical risks are closely interconnected. They highlight how faculty perceptions are shaped by the challenges of maintaining assessment credibility while adapting teaching practices in AI-mediated environments, rather than reflecting isolated or independent issues. Equity, ethics, and institutional responsibility While the ethical and equity challenges of generative AI have been explored in global contexts, their specific manifestations in higher education remain underexamined. Equity and ethics complicate the integration of GenAI in academic settings. UNESCO18 emphasises the need for human-centered, transparent, and accountable use of AI, highlighting the importance of clear ethical governance. A key concern relates to unequal access to AI tools. Watson and Romic39 show that access to advanced AI technologies, particularly paid services, shapes how students benefit from these tools. Sullivan et al.40 further note that such imbalances may translate into academic advantages for some students. Differences in access and familiarity reflect broader digital inequalities that influence how students engage with academic tasks. More experienced users can leverage AI strategically, while others remain disadvantaged, potentially reinforcing existing educational inequalities41. Beyond issues of access, AI-generated content challenges the validity of traditional assessment practices. Cotton et al.20 demonstrate that it is often difficult to distinguish between student work and machine-generated output, raising fundamental concerns about how assessment can remain valid and fair in AI-mediated environments. In response to these challenges, institutional approaches have begun to emerge. For example, a previous study examined the use of ChatGPT in public universities in Hong Kong, highlighting institutional restrictions and regulatory measures governing its use in academic settings42. Ethical concerns extend further to data privacy and algorithmic bias. These issues are embedded in how AI systems operate and are often invisible to users. Algorithmic bias can shape what knowledge is produced and whose perspectives are marginalised, reinforcing epistemic inequalities. Gao et al.30 describe this as “structural epistemic injustice,” driven by biased data representation, data quality issues, and the authority granted to data-driven outcomes. Privacy concerns also affect academic practice. They relate directly to autonomy and institutional trust, as faculty and students may hesitate to engage with systems that collect and analyse their data without transparency. Kayas et al.43 show that privacy in higher education is socially negotiated and shapes how academics construct their professional identity. These findings point to interconnected challenges at the intersection of equity, ethics, and governance in higher education. These challenges influence how AI is understood, trusted, and used in academic practice. Further context-specific evidence is needed to understand how they are experienced within national contexts, particularly in Saudi higher education. Institutional responses and faculty capacity building In response to the challenges posed by GenAI, higher education institutions have adopted diverse policy and governance approaches. Some institutions have introduced restrictive policies governing AI use in assessment36,44. Recent research also underscores the importance of developing structured AI competencies among educators to support the effective integration of artificial intelligence into teaching practices45. A recurring theme in the literature is the need for targeted professional development to enhance faculty AI literacy. Studies point to gaps in faculty understanding of GenAI capabilities25, its limitations26, and its ethical implications46. In this context, recent work emphasizes the importance of equipping academic staff with the knowledge and skills required to engage with GenAI in both teaching and assessment practices44. These responses indicate that GenAI is not only a technological issue but also an institutional and pedagogical challenge. Effective integration requires clear guidance and professional development to support faculty in applying GenAI appropriately in teaching practice. Saudi-specific context Despite the expanding international literature on the implications of generative artificial intelligence (GenAI) in higher education, research in the Saudi context has primarily focused on AI adoption, technological readiness, and student acceptance, with comparatively limited attention to faculty perceptions of tools such as ChatGPT and their implications for assessment, academic integrity, and institutional governance. Existing studies in Saudi universities provide valuable but partial insights into the integration of artificial intelligence in higher education. Some research emphasizes students’ engagement with tools such as ChatGPT47 and explores broader opportunities and challenges associated with AI adoption48, while other studies adopt policy-oriented or technical perspectives, focusing on national AI strategies49 or specific applications such as educational robotics50. More recent institutional research has examined readiness and implementation requirements, highlighting the role of both technical and human factors in shaping AI adoption51. In a related context, a previous study examined the use of artificial intelligence techniques in teacher preparation programs at a Saudi university from the perspective of faculty members52. A recent study identified significant barriers to the use of AI in Blackboard, particularly related to insufficient training and institutional support, while also highlighting the influence of demographic factors on adoption53. Complementing these findings, a recent systematic review underscores the presence of key ethical challenges associated with the use of AI in higher education, particularly in relation to academic integrity, privacy, fairness, and transparency54. In parallel, descriptive evidence from the Saudi context indicates growing awareness of the importance of artificial intelligence in education, alongside the need for teacher training, clear implementation standards, and alignment with broader socio-cultural and institutional factors55. Existing studies indicate that AI integration in Saudi higher education is frequently examined through segmented lenses—such as student use, technical infrastructure, or institutional preparedness—while providing comparatively limited insight into how faculty interpret and navigate generative AI within everyday teaching and assessment practices. Accordingly, the relationship between faculty perceptions, academic integrity, and assessment practices remains only partially addressed within the Saudi context. Research gaps Research on the integration of generative artificial intelligence (GenAI) in higher education has examined key dimensions, including faculty attitudes, academic integrity, assessment practices, and institutional responses across diverse international contexts. While this body of work provides important insights into the opportunities and challenges associated with tools such as ChatGPT, it has often approached these dimensions in relative isolation. However, in practice, the use of generative AI in higher education brings these elements into closer interaction. Concerns about academic integrity are directly linked to assessment design, while institutional policies shape how faculty interpret, regulate, and engage with AI tools. This interdependence suggests the need for a more integrated analytical perspective that captures how these dimensions are experienced and negotiated within academic practice. Within the Saudi context, this pattern can also be observed, where existing research tends to emphasize adoption and readiness, while offering comparatively limited insight into how faculty members interpret and respond to generative AI in relation to core academic processes. Accordingly, this study contributes to the literature by providing empirically grounded, context-sensitive insights into how faculty members in Saudi universities engage with ChatGPT, with particular attention to the interconnections between academic integrity, assessment practices, and institutional roles. Methodology This study examines faculty members’ perceptions of the use of ChatGPT in Saudi higher education, with a particular focus on challenges related to academic integrity, assessment practices, and pedagogical use, as well as proposed solutions for its responsible integration. Ethical approval Ethical approval for this study was obtained from the Local Research Ethics Committee for Research Involving Living Creatures at the University of Hafr Al-Batin. The research protocol, survey instrument, and data collection procedures were reviewed and approved in accordance with institutional research regulations and internationally recognized ethical standards for research involving human participants. Approval was granted under Approval No. HPO-05-FT-25/18 in June 2024, prior to the commencement of data collection. The scope of the approval covered participant recruitment, data collection procedures, confidentiality safeguards, and data management practices. The study did not involve any clinical experimentation. All methods were performed in accordance with the relevant institutional guidelines and regulations and with the 1964 Declaration of Helsinki and its later amendments. Research design This study adopted a sequential explanatory mixed-methods design to provide a comprehensive understanding of faculty perceptions of the use of ChatGPT in higher education. In this design, quantitative data were collected and analyzed first to identify general patterns, followed by qualitative data to explain and elaborate on the quantitative results56,57. This methodological design was selected to align with the study objectives. Specifically, the quantitative phase addressed the objective of identifying overall patterns and group differences in faculty perceptions of academic integrity and institutional roles related to ChatGPT, whereas the qualitative phase supported the objective of providing deeper explanations for these patterns through participants’ perspectives. Integrating both phases enabled a more comprehensive understanding of the research problem than either method could provide independently. The quantitative component was used to identify general patterns and differences across groups in faculty perceptions of challenges and proposed solutions across academic rank, years of experience, and institutional roles. This phase provided a broad and generalizable overview of how faculty members perceive the implications of ChatGPT in higher education. The qualitative component complemented the quantitative findings by offering deeper insights into participants’ perceptions, concerns, and suggested responses. Through the analysis of open-ended responses, this phase enabled a more detailed understanding of how faculty interpret and make sense of the challenges and solutions associated with the use of ChatGPT. In this study, the focus is specifically on faculty perceptions rather than experiences, as the aim is to examine how participants evaluate and interpret the implications of ChatGPT within their academic context. The integration of the quantitative and qualitative components was achieved at the interpretation stage, where qualitative findings were used to explain and contextualize the statistical patterns identified in the quantitative analysis. This integration enhanced the overall explanatory power of the study and provided a more comprehensive understanding of the research problem. Participants and instrument Participants in this study were recruited using a non-probability convenience sampling approach. Faculty members and academic leaders were invited through institutional email networks and academic communication channels, ensuring that participants had direct engagement with teaching, assessment, and academic decision-making processes. This approach is appropriate in perception-based studies where participants are selected based on their relevance to the research topic56. Efforts were made to reach faculty members across universities located in different regions of Saudi Arabia, including the central, eastern, western, northern, and southern regions. This distribution was intended to capture variation in institutional contexts and academic environments. The resulting sample reflects a broad range of perspectives across diverse geographical and institutional settings within Saudi higher education. A total of 300 faculty members and academic leaders were invited to participate via email. The invitation included information about the purpose of the study, its academic significance, and the estimated completion time, as well as ethical assurances regarding voluntary participation, confidentiality, and anonymity. Participants were informed that proceeding with the survey constituted informed consent. This sample size was selected to ensure sufficient participation and variability in responses for the purposes of the study. In total, 202 complete and valid responses were received (response rate = 67.3%), while incomplete responses were excluded from the analysis. To assess the potential for non-response bias, an early–late response analysis was conducted, which is widely regarded as a suitable proxy in survey-based research. Respondents were divided into two groups based on their response timing: early respondents (those who completed the survey following the initial invitation) and late respondents (those who responded after follow-up reminders). The underlying assumption is that late respondents are more similar to non-respondents; therefore, any systematic differences between the two groups may indicate the presence of non-response bias. Independent-samples t-tests were performed to compare the groups across all key study variables. The results revealed no statistically significant differences between early and late respondents, suggesting that non-response bias is unlikely to threaten the validity of the study’s findings. For the qualitative component, open-ended responses were collected as part of the survey to provide complementary insights. These qualitative data were used to enrich and contextualize the quantitative findings. Data collection was conducted between July 2024 and February 2025 using an online questionnaire administered through Google Forms. Ethical approval for the study was obtained through appropriate institutional procedures prior to data collection. The questionnaire was designed to measure faculty perceptions of challenges associated with the use of ChatGPT and proposed solutions for its responsible integration in higher education. Structured Likert-scale items were used to quantify participants’ perceptions, as such instruments are widely employed in educational research to assess attitudes and perceptions in a systematic and comparable manner. The questionnaire consisted of two main sections designed to capture faculty perceptions of ChatGPT in higher education. The first section included 11 Likert-scale items measuring perceived challenges associated with the use of ChatGPT, covering issues related to academic integrity, assessment validity, and pedagogical concerns. The second section comprised 9 Likert-scale items assessing proposed solutions for the responsible integration of ChatGPT in higher education. These items were treated as two conceptually distinct dimensions (challenges and solutions), and the questionnaire was not structured into formal sub-scales. A three-point Likert scale (Agree, Neutral, Disagree) was used to capture broad patterns of faculty perceptions rather than subtle attitudinal differences. While this format provides less sensitivity than five- or seven-point scales, it has been widely used in prior research and remains appropriate for studies examining general response patterns58–61. Given the descriptive focus of the study, the diversity of participants’ academic and administrative roles, and the aim of identifying overall patterns of perception rather than fine-grained differences, the three-point scale was considered appropriate. Its limitations, including reduced response variability, were acknowledged when interpreting the findings. The development of the questionnaire was informed by a review of international and Saudi-based literature on artificial intelligence in higher education, ensuring contextual relevance. Prior to data collection, the instrument underwent an initial review phase involving a small group of academic professionals (n = 7) with relevant expertise in higher education and AI-related practices. Their evaluation focused on the clarity of item wording, conceptual relevance, and overall appropriateness of the measures. Feedback from these reviewers was systematically incorporated through iterative revisions to improve item clarity, eliminate ambiguity, and enhance the overall structure of the instrument. The questionnaire captures a range of context-specific challenges and proposed solutions, with each item representing a distinct aspect of the phenomenon. Given this multidimensional focus, the items are not assumed to reflect a single underlying factor; accordingly, factor analysis was not central to the objectives of the study. Cronbach’s alpha for the challenges items was high (α = 0.88). The proposed solutions items also demonstrated acceptable reliability (α = 0.71). To complement the quantitative data, two open-ended questions were included to capture participants’ perceptions, concerns, and suggested responses in their own words. In mixed methods designs, such open-ended survey items are commonly used to provide explanatory depth and contextual interpretation of quantitative findings, particularly when the qualitative component is intended to support rather than replace in-depth qualitative inquiry56,57. These responses were analyzed using thematic analysis, following Braun and Clarke’s62 approach. The analysis involved familiarization with the data, initial coding, and the identification and refinement of themes, ensuring a systematic and transparent interpretation of the qualitative findings. Sample characteristics As shown in Table 1, the sample demonstrated diversity across academic rank, years of experience, and administrative position. Assistant Professors constituted the largest group (n = 82), followed by Associate Professors (n = 72) and Professors (n = 48), indicating a relatively balanced distribution across academic ranks and suggesting that the sample captures perspectives from different levels of academic seniority. Table 1Demographic characteristics of the sample.VariableCategoryFrequency (N)Percent (%)*Academic rankProfessor4824%Assistant Professor8241%Associate Professor7236%Years of experienceLess than 5 years4221%More than 10 years9648%5–10 years6432%PositionDepartment Head178%Faculty Member8944%College Dean4020%University Vice President2412%College Vice Dean3216%*Percentages may not total 100% due to rounding. Demographic characteristics of the sample. *Percentages may not total 100% due to rounding. In terms of professional experience, a substantial proportion of participants reported more than ten years of experience (n = 96), while those with five to ten years (n = 64) and less than five years (n = 42) were less represented. This pattern suggests that the findings are primarily informed by experienced faculty members, which may influence the interpretation of AI-related challenges and solutions. Regarding administrative roles, faculty members without administrative positions formed the largest category (n = 89), followed by College Deans (n = 40), Vice Deans (n = 32), University Vice Presidents (n = 24), and Department Heads (n = 17). This distribution may reflect a typical academic structure and suggests that the results are more strongly influenced by teaching-focused faculty perspectives than by administrative viewpoints. Gender distribution was not included as a variable in this study, as the focus was on faculty perceptions regardless of gender. All participants were drawn from public universities in Saudi Arabia. Data analysis Quantitative analysis Quantitative data were analyzed using descriptive statistics, including means, standard deviations, frequencies, and percentages, to summarize responses to closed-ended items. Although responses were collected using a three-point ordinal Likert scale, means and standard deviations were reported solely for descriptive purposes to facilitate comparison across items, whereas inferential analyses relied on non-parametric procedures appropriate for ordinal data. Given the ordinal nature of the 3-point Likert-scale data (1 = Disagree, 2 = Neutral, 3 = Agree) and evidence of non-normal distribution, non-parametric statistical tests were employed. Specifically, the Kruskal–Wallis H test was used to examine differences in faculty perceptions of challenges and proposed solutions across academic rank, years of experience, and administrative position. Where statistically significant differences were identified, Dunn–Bonferroni post-hoc pairwise comparisons were conducted to determine specific group differences. In addition, effect sizes (ηH2) with 95% confidence intervals were calculated to assess the magnitude of observed differences, allowing interpretation of both statistical and practical significance. Statistical significance was set at p < .05. Table 2 presents descriptive statistics for faculty perceptions of challenges associated with the use of ChatGPT in Saudi higher education (N = 202). The results indicate generally elevated levels of concern across several dimensions, particularly those related to academic integrity and assessment practices. Table 2Faculty perceptions of the challenges of using ChatGPT in Saudi higher education.NoItemsAgreeNeutralDisagreeMSD%1Using ChatGPT may generate biased information113(56%)65(32%)24(12%)2.440.702Using ChatGPT makes it easier for students to cheat178(88%)16(8%)8(4%)2.840.463Using ChatGPT may lead to unequal evaluation176(87%)20(10%)6(3%)2.840.444ChatGPT is used in ways that are inconsistent with human values145(72%)32(16%)25(12%)2.600.705ChatGPT works on plagiarism and fake citations145(72%)32(16%)25(12%)2.600.706Overreliance on ChatGPT may negatively impact critical thinking and problem-solving skills170(84%)16(8%)16(8%)2.760.587Reliance on ChatGPT programs may lead to poor academic performance178(88%)4(2%)20(10%)2.780.618Using ChatGPT may limit students’ creativity145(72%)32(16%)25(12%)2.600.709Using ChatGPT may lead to a breach of confidentiality and data privacy129(64%)57(28%)16(8%)2.560.6410Using ChatGPT may lead to poor scientific research skills137(68%)32(16%)33(16%)2.510.7611Using ChatGPT may lead to a lack of human interaction between students and faculty members145(72%)40(20%)17(8%)2.630.63 Faculty perceptions of the challenges of using ChatGPT in Saudi higher education. The highest mean scores were observed for items related to cheating (M = 2.84) and unequal evaluation (M = 2.84), suggesting that faculty concerns are primarily centered on the implications of ChatGPT for assessment validity and fairness. These findings highlight the perceived difficulty of verifying authorship and ensuring reliable evaluation in AI-assisted academic contexts. Concerns regarding the impact of ChatGPT on critical thinking (M = 2.76) and academic performance (M = 2.78) further indicate apprehension about its potential influence on students’ independent learning processes and higher-order cognitive development. Other dimensions, including bias, ethical alignment, and plagiarism-related risks, showed elevated mean scores (M ranging from 2.44 to 2.60), suggesting broader pedagogical and ethical concerns surrounding generative AI use. In contrast, lower mean scores for data privacy (M = 2.56) and reduced human interaction (M = 2.63), relative to other dimensions, suggest that faculty members may prioritize immediate academic and assessment-related risks over longer-term or indirect implications. Overall, the pattern of responses appears to suggest a consistent emphasis on maintaining academic integrity and preserving the credibility of assessment practices in the context of generative AI. Table 3 presents descriptive statistics for faculty perceptions of proposed solutions to address the challenges associated with the use of ChatGPT in Saudi higher education (N = 202). The results show elevated mean scores across most proposed interventions (M ranging from 2.56 to 2.96). Table 3Faculty perceptions of proposed solutions to address the challenges of ChatGPT in Saudi higher education.NoItemsAgreeNeutralDisagreeMSD%1Spread awareness of ChatGPT concepts and practices to students and faculty186(92%)12 (6%)4(2%)2.900.192Training faculty members on the use of artificial intelligence programs and tools186(92%)12(6%)4(2%)2.900.193Developing educational curricula in line with the artificial intelligence revolution162(80%)24(12%)16(8%)2.720.434Determine the percentages of permissible use of artificial intelligence in assignments and research161(80%)25(12%)16(16%)2.720.435Applying in-person tests on campus for all courses162(80%)16(8%)24(12%)2.680.466Motivating students to do research and assignments in the university library146(72%)24(12%)32(16%)2.560.527Request assessments from students based on participation in group discussions, presentations, or other interactive activities194(96%)8(4%)0%2.960.208Establish clear guidelines in university regulations on how to use artificial intelligence177(88%)17(8%)8(4%)2.840.259Using plagiarism detection tools to evaluate submitted assignments and research185(92%)17(8%)0%2.910.28 Faculty perceptions of proposed solutions to address the challenges of ChatGPT in Saudi higher education. The highest level of agreement was observed for interactive and participation-based assessment methods (M = 2.96), suggesting a strong preference for pedagogical approaches that promote authentic student engagement and reduce reliance on AI-generated content. This may reflect faculty recognition of the need for assessment practices that better capture individual student performance in AI-mediated contexts. Similarly, awareness initiatives and faculty training (M = 2.90) received substantial support, indicating that capacity building is viewed as a critical foundation for the responsible integration of artificial intelligence in higher education. These findings suggest that faculty perceive AI-related challenges as extending beyond technical issues to include pedagogical and ethical dimensions. Institutional measures, including the development of AI-aligned curricula (M = 2.72) and the establishment of clear regulatory guidelines (M = 2.84), also received strong endorsement. This highlights the perceived importance of coordinated institutional responses and policy frameworks in addressing the complexities of generative AI adoption. In contrast, lower mean scores were observed for more traditional strategies, such as encouraging library-based work (M = 2.56), suggesting that faculty may view conventional approaches as insufficient in addressing the evolving challenges associated with AI technologies. Overall, the findings reflect a clear orientation toward structured, policy-driven, and pedagogically adaptive solutions in response to the growing presence of generative AI in higher education. The purpose of this study was to examine whether faculty members’ perceptions of the challenges associated with using ChatGPT in academic settings differ according to their academic rank (Full Professor, Associate Professor, Assistant Professor). The Kruskal–Wallis test results (Table 4) revealed statistically significant differences across all eleven challenge variables, with effect sizes (ηH2) ranging from moderate to large. Post-hoc pairwise comparisons using Dunn’s test (Table 5) further clarified the nature of these rank-based differences. Table 4Differences in perceptions of the challenges associated with using ChatGPT according to academic rank (Kruskal–Wallis Test results and effect sizes).Challenge variableAcademic rankMeanHp-valueηH2 (95% CI)MagnitudeUsing ChatGPT may generate biased informationFull Professor2.1725.8 < .0010.119 [0.05, 0.21]ModerateUsing ChatGPT makes it easier for students to cheatAssistant Prof2.33Using ChatGPT may lead to unequal evaluationAssociate Prof2.78ChatGPT is used in ways that are inconsistent with human valuesFull Professor2.3386.6 < .0010.425 [0.30, 0.60]LargeChatGPT works on plagiarism and fake citationsAssistant Prof3Overreliance on ChatGPT may negatively impact critical thinking and problem-solving skillsAssociate Prof3Reliance on ChatGPT programs may lead to poor academic performanceFull Professor2.1791.1 < .0010.448 [0.34, 0.61]LargeUsing ChatGPT may limit students’ creativityAssistant Prof3.00Using ChatGPT may lead to a breach of confidentiality and data privacyAssociate Prof2.89Using ChatGPT may lead to poor scientific research skillsFull Professor2.3321.9 < .0010.100 [0.03, 0.22]ModerateUsing ChatGPT may lead to a lack of human interaction between students and faculty membersAssistant Prof2.49Associate Prof2.89CH5: ChatGPT works on plagiarism and fake citationsFull Professor2.1726.4 < .0010.123 [0.03, 0.24]ModerateAssistant Prof2.58Associate Prof2.89CH6: Overreliance on ChatGPT may negatively impact critical thinking skillsFull Professor2.528.3 < .0010.132 [0.07, 0.22]ModerateAssistant Prof3.00Associate Prof2.67CH8: Reliance on ChatGPT programs may lead to poor academic performanceFull Professor2.1761 < .0010.297 [0.16, 0.44]LargeAssistant Prof2.90Associate Prof3.008. Using ChatGPT may limit students’ creativityFull Professor2.1756.8 < .0010.275 [0.20, 0.39]LargeAssistant Prof2.49Associate Prof3.00CH9: Using ChatGPT may lead to a breach of confidentiality and data privacyFull Professor2.0038.1 < .0010.181 [0.08, 0.33]LargeAssistant Prof2.69Associate Prof2.78CH10: Using ChatGPT may lead to poor scientific research skillsFull Professor2.1714.7.0010.064 [0.01, 0.18]ModerateAssistant Prof2.47Associate Prof2.78CH11: Using ChatGPT may lead to a lack of human interactionFull Professor2.6719.4 < .0010.088 [0.04, 0.19]ModerateAssistant Prof2.37Associate Prof2.89 Differences in perceptions of the challenges associated with using ChatGPT according to academic rank (Kruskal–Wallis Test results and effect sizes). Table 5Post-hoc pairwise comparisons (Dunn’s Test).VariableGroup 1Group 2z-statisticAdjusted pSignifCH1AssistantAssociate3.56.001**AssociateFull-4.84 < .001****CH2AssistantFull-8.47 < .001****AssociateFull-8.26 < .001****CH3AssistantFull-9.24 < .001****AssociateFull-7.58 < .001****CH4AssistantAssociate2.92.011*AssociateFull-4.59 < .001****CH5AssistantAssociate2.68.022*AssistantFull-2.87.012*AssociateFull-5.13 < .001****CH6AssistantAssociate-3.75 < .001***AssistantFull-4.99 < .001****CH7AssistantFull-6.34 < .001****AssociateFull-7.50 < .001****CH8AssistantAssociate4.25 < .001****AssistantFull-3.89 < .001***AssociateFull-7.48 < .001****CH9AssistantFull-5.22 < .001****AssociateFull-5.80 < .001****CH10AssistantFull-2.41.048*AssociateFull-3.84 < .001***CH11AssistantAssociate4.39 < .001****Significance levels are: ****p < .001; ***p < .001; **p < .01; *p < .05. Post-hoc pairwise comparisons (Dunn’s Test). Significance levels are: ****p < .001; ***p < .001; **p < .01; *p < .05. Notably, the largest effect sizes were observed for challenges related to academic integrity and evaluation fairness. Specifically, “Using ChatGPT makes it easier for students to cheat” (ηH2 = 0.425, large) and “Using ChatGPT may lead to unequal evaluation” (ηH2 = 0.448, large) yielded the highest H-values. Post-hoc comparisons indicated that Full Professors (Mean = 2.33 for cheating; 2.17 for unequal evaluation) held significantly lower concern levels compared to both Assistant and Associate Professors (p < .001 for all comparisons). This suggests that junior faculty may perceive a more immediate threat to classroom assessment practices, possibly due to their closer, daily instructional engagement with students. Similarly, “Reliance on ChatGPT may lead to poor academic performance” (ηH2 = 0.297, large) and “Using ChatGPT may limit students’ creativity” (ηH2 = 0.275, large) showed large magnitude differences. For both variables, Full Professors again reported the lowest mean scores (2.17 for poor performance; 2.17 for limiting creativity), while Associate Professors consistently reported the highest (3.00 for both). Dunn’s test confirmed that Full Professors differed significantly from both Assistant and Associate Professors (p < .001 for most comparisons). This pattern may suggest that senior faculty might view generative AI as a supplementary tool rather than a fundamental impediment to skill development, whereas mid-career and junior faculty express stronger apprehensions about long-term academic skill erosion. Moderate effect sizes (ηH2 between 0.064 and 0.181) were found for variables such as “Using ChatGPT may generate biased information” (ηH2 = 0.119), “ChatGPT is used in ways inconsistent with human values” (ηH2 = 0.100), “ChatGPT works on plagiarism and fake citations” (ηH2 = 0.123), “Overreliance may negatively impact critical thinking” (ηH2 = 0.132), and “May lead to poor scientific research skills” (ηH2 = 0.064). For most of these, Associate Professors again scored highest (e.g., 2.89 for inconsistent with human values; 2.89 for plagiarism/fake citations), while Full Professors scored lowest. Interestingly, for “lack of human interaction” (ηH2 = 0.088, moderate), Assistant Professors recorded the lowest concern (Mean = 2.37) compared to Associate (2.89) and Full (2.67), suggesting that junior faculty may be more comfortable with AI-mediated interactions or view them as a necessary evolution of digital learning environments. Across all eleven variables, two robust patterns emerged from Table 3. The most frequent and strongest significant differences occurred between Full Professors and the other two ranks (e.g., CH2, CH3, CH7, CH9, all with p < .001). Full Professors consistently reported lower mean concern scores than both junior ranks. With respect to the compressions between Assistant vs. Associate Professors fewer pairwise differences reached significance. Notable exceptions included CH1 (biased information, p = .001), CH4 (inconsistent with human values, p = .011), CH8 (limiting creativity, p < .001), and CH11 (lack of human interaction, p < .001). In each of these, Assistant Professors reported lower means than Associate Professors, indicating that mid-career faculty may be the most sensitive to several ethical and pedagogical risks. These findings collectively suggest that academic rank serves as a meaningful proxy for differing exposure, responsibility, and pedagogical experience with AI tools. Full Professors, who often focus on higher-level research and institutional strategy, may perceive ChatGPT’s challenges as less severe—perhaps viewing them as manageable or overstated. In contrast, Assistant Professors, who are typically deeply involved in frontline teaching and grading, express heightened concerns about cheating, evaluation inequality, and poor academic performance. Associate Professors frequently exhibited the highest concern levels, possibly reflecting a transitional role where they balance teaching oversight with emerging administrative responsibilities, making them acutely aware of both student behaviors and policy gaps. As shown in Tables 6 and 7, the analysis of proposed solutions for integrating ChatGPT into higher education reveals significant differences in perception across academic ranks for nearly all items, except for the development of curricula (SOL3). The findings suggest a clear generational or hierarchical divide in how faculty members prioritize institutional interventions. Table 6Differences in perceptions of the solutions associated with using ChatGPT according to academic rank.SolutionAcademic RankMeanHp-valueηH2 (95% CI)MagnitudeSOL1: Spread awareness of ChatGPT concepts and practices to students and facultyFull Professor2.3313.0010.055 [0.03, 0.11]SmallAssistant Prof3Associate Prof3SOL12: Training faculty members on using AI programs and toolsFull Professor2.6755.5 < .0010.269 [0.14, 0.45]LargeAssistant Prof3Associate Prof3SOL3: Developing curricula in line with AI advancementsFull Professor2.830.955.62-0.005 [-0.01, 0.04]SmallAssistant Prof2.69Associate Prof2.67SOL4: Determining permissible use of AI in assignments and researchFull Professor2.528.3 < .0010.132 [0.08, 0.20]ModerateAssistant Prof2.58Associate Prof3SOL5: Applying in-person tests on campus for all coursesFull Professor2.3329.7 < .0010.139 [0.04, 0.27]ModerateAssistant Prof2.79Associate Prof2.78SOL6: Motivating students to use the university library for assignmentsFull Professor245.3 < .0010.218 [0.09, 0.36]LargeAssistant Prof2.79Associate Prof2.67SOL7: Requesting assessments based on interactive activitiesFull Professor2.8326.6 < .0010.124 [0.04, 0.21]ModerateAssistant Prof3Associate Prof3SOL8: Establishing guidelines for AI use in university regulationsFull Professor2.8316.4 < .0010.072 [0.04, 0.13]ModerateAssistant Prof2.68Associate Prof3SOL9: Using plagiarism detection tools in evaluating assignmentsFull Professor2.8311.5.0030.048 [0.02, 0.12]SmallAssistant Prof2.88Associate Prof3 Differences in perceptions of the solutions associated with using ChatGPT according to academic rank. Table 7Post-hoc pairwise comparisons (Dunn’s Test) for solutions.VariableGroup 1Group 2z-statisticAdjusted pSignifSOL1AssistantAssociate-2.54.033*AssistantFull-3.39.002**AssociateFull-1.10.812nsSOL2AssistantFull-6.77 < .001****AssociateFull-6.61 <.001****AssistantAssociate0.001.000nsSOL4AssistantAssociate4.59 < .001****AssociateFull-4.51 < .001****AssistantFull-0.551.000nsSOL5AssistantFull-5.04 < .001****AssociateFull-4.72 < .001****AssistantAssociate-0.231.000nsSOL6AssistantFull-6.59 < .001****AssociateFull-5.14 < .001****AssistantAssociate-1.48.416nsSOL7AssistantFull-4.69 < .001****AssociateFull-4.57 < .001****AssistantAssociate0.001.000nsSOL8AssistantAssociate3.95 < .001***AssociateFull-2.60.028*AssistantFull0.851.000nsSOL9AssistantAssociate2.44.044*AssociateFull-3.21.004**AssistantFull-1.13.782nsp-values are adjusted using the Bonferroni correction. Only variables with a significant overall Kruskal–Wallis result (p < .05) are detailed here. SOL3 was omitted as it showed no significant group differences. Post-hoc pairwise comparisons (Dunn’s Test) for solutions. p-values are adjusted using the Bonferroni correction. Only variables with a significant overall Kruskal–Wallis result (p < .05) are detailed here. SOL3 was omitted as it showed no significant group differences. A primary trend across the results is the consistent divergence of Full Professors from their junior colleagues. For solutions such as faculty training (SOL2), library utilization (SOL6), and interactive assessments (SOL7), Full Professors reported significantly lower mean scores than Assistant and Associate Professors. This is most evident in SOL2 (Training faculty), where the effect size was large (eta2_H = 0.269). Dunn’s post-hoc test confirms that Full Professors differ significantly (p < .001) from both lower ranks, who showed a unanimous high preference (Mean = 3.00) for training. This may suggest that senior faculty, perhaps due to established pedagogical routines or a different stage of career focus, perceive a lower personal or institutional necessity for specialized AI training compared to early-career faculty who face longer-term integration of these tools. Interestingly, Associate Professors emerged as the group most supportive of formal regulation and detection. For SOL8 (Establishing guidelines) and SOL9 (Plagiarism detection tools), Associate Professors maintained a perfect mean of 3.00, which was significantly higher than Assistant Professors. In the case of SOL8, the pairwise comparison (z = 3.95, p < .001) suggests that mid-career faculty may be more sensitive to the administrative and legal vacuums created by AI, viewing institutional regulations as a necessary safeguard for academic integrity. As for areas of institutional consensus and friction, Curriculum Development (SOL3) was the only variable showing no significant difference (p = .620) and a negligible effect size. This represents a "point of consensus" where all ranks—regardless of seniority—recognize that the core solution lies in updating the curriculum to match AI advancements. With respect to In-Person Testing (SOL5) there was a moderate effect size (eta^2_H = 0.139). However, the post-hoc analysis shows that Assistant and Associate Professors are aligned in their higher support for returning to on-campus, in-person testing to mitigate AI cheating, significantly differing from Full Professors (p < .001). Library Use (SOL6) was one of the largest disparities was found in the motivation to use the university library. Full Professors showed the lowest interest (Mean = 2.00) compared to Assistant Professors (Mean = 2.79). This indicates that junior faculty may see traditional research environments as a more viable “safe zone” against AI overreliance than senior faculty do. The data suggest that there is broad agreement on the need for curricular reform, while support for implementation-focused solutions—such as training, detection tools, and interactive assessments—is reported more strongly among Assistant and Associate Professors. The large effect sizes observed for SOL2 and SOL6 indicate that differences across academic ranks should be considered in institutional policy design, as uniform approaches to faculty AI readiness may not fully address variation in support across groups. As shown in Tables 8 and 9, the results indicate that professional experience significantly shapes how faculty perceive the challenges of ChatGPT integration. While some concerns are universal, the intensity of these fears often diminishes with increased years of experience. Table 8Differences in Perceptions of Challenges Associated with ChatGPT Based on Years of Experience.VariableYears of experienceMeanHp-valueηH295% CIMagnitudeCH1: Biased infoLess than 5 years2.268.93.0120.035[-0.001, 0.12]Small5–10 years2.63More than 10 years2.42CH2: Easier to cheatLess than 5 years3.0029.8 < .0010.140[0.08, 0.20]Moderate5–10 years3.00More than 10 years2.67CH3: Unequal evaluationLess than 5 years3.0054.4 < .0010.263[0.18, 0.36]Large5–10 years3.00More than 10 years2.50CH4: Inconsistent valuesLess than 5 years2.5812.0.0020.050[0.001, 0.16]Small5–10 years2.75More than 10 years2.50CH5: Plagiarism/FakesLess than 5 years2.3413.9 < .0010.060[0.01, 0.14]Small5–10 years2.88More than 10 years2.50CH6: Critical thinkingLess than 5 years3.0013.2.0010.056[0.03, 0.13]Small5–10 years2.75More than 10 years2.67CH7: Poor performanceLess than 5 years3.0015.6 < .0010.068[0.03, 0.13]Moderate5–10 years2.88More than 10 years2.58CH8: Limit creativityLess than 5 years2.581.53.4650.000[-0.01, 0.05]Small5–10 years2.63More than 10 years2.58CH9: Privacy breachLess than 5 years2.798.83.0120.034[0.001, 0.11]Small5–10 years2.63More than 10 years2.42CH10: Research skillsLess than 5 years2.342.67.2630.003[-0.01, 0.06]Small5–10 years2.63More than 10 years2.50CH11: Interaction lackLess than 5 years2.345.53.0630.018[-0.01, 0.10]Small5–10 years2.63More than 10 years2.75 Differences in Perceptions of Challenges Associated with ChatGPT Based on Years of Experience. Table 9Post-hoc pairwise comparisons (Dunn’s Test) for experience.VariableGroup 1Group 2z-statisticAdjusted pSignifCH1 < 5 yrs5–10 yrs2.67.023*5–10 yrs > 10 yrs-2.45.043*CH2 < 5 yrs > 10 yrs-4.16 < .001****5–10 yrs > 10 yrs-4.77 < .001****CH3 < 5 yrs > 10 yrs-5.62 < .001****5–10 yrs > 10 yrs-6.44 < .001****CH45–10 yrs > 10 yrs-3.35.002**CH5 < 5 yrs5–10 yrs3.31.003**5–10 yrs > 10 yrs-3.13.005**CH6 < 5 yrs > 10 yrs-3.58.001**CH7 < 5 yrs > 10 yrs-3.78 < .001***CH9 < 5 yrs > 10 yrs-2.96.009**p-values are adjusted using the Bonferroni correction. Comparisons are shown for variables with significant overall differences (p < .05). Post-hoc pairwise comparisons (Dunn’s Test) for experience. p-values are adjusted using the Bonferroni correction. Comparisons are shown for variables with significant overall differences (p < .05). A compelling trend is observed in CH2 (Cheating) and CH3 (Unequal Evaluation). Faculty with less than 10 years of experience (both the < 5 and 5–10 year groups) expressed the highest possible level of concern (Mean = 3.00), while faculty with more than 10 years of experience were significantly less concerned (p < .001). The large effect size for Unequal Evaluation (eta^2_H = 0.263) suggests that junior and mid-career faculty may feel more vulnerable to the disruptions AI causes in grading and academic fairness than their more senior counterparts. Faculty with less than 5 years of experience are significantly more worried about the impact of AI on student skills. For CH6 (Critical Thinking) and CH7 (Academic Performance), the most junior faculty reported maximum concern (Mean = 3.00). This sensitivity may be related to these instructors being more closely involved in foundational teaching, where the development of these core skills is most visible and at risk. Notably, CH8 (Students’ creativity) and CH10 (Scientific research skills) showed no significant differences across the three groups (p > .05). This indicates a cross-generational academic consensus: regardless of how long a faculty member has been teaching, they share a moderate, consistent concern that AI may hinder the creative and rigorous nature of scientific inquiry. Concerns regarding CH9 (Data Privacy) and CH1 (Biased Information) followed a downward trend as experience increased. Younger faculty (< 5 years) are the most concerned about privacy (Mean = 2.79), whereas the mid-career group (5–10 years) is the most concerned about the accuracy and bias of the information AI produces (Mean = 2.63). This suggests that different career stages prioritize different technical risks, with junior faculty focusing on security and mid-career faculty focusing on content quality. As shown in Tables 10 and 11, professional experience was associated with significant differences in faculty preferences for several AI-related interventions. However, these differences were not uniform across all proposed solutions. Table 10Differences in perceptions of the solutions associated with using ChatGPT According to experience.SolutionYears of experienceMeanHp-valueηH295% CIMagnitudeSOL1: Spread awarenessLess than 5 years3.005.45.0660.017[0.007, 0.07]Small5–10 years2.92More than 10 years2.88SOL2: Training facultyLess than 5 years3.0019.1 <.0010.086[0.05, 0.14]Moderate5–10 years2.83More than 10 years3.00SOL3: Curricula updateLess than 5 years3.0057.5 < .0010.279[0.16, 0.42]Large5–10 years2.92More than 10 years2.25SOL4: Permissible useLess than 5 years2.761.70.4270.000[-0.01, 0.06]Small5–10 years2.75More than 10 years2.63SOL5: In-person testsLess than 5 years3.0013.1.0010.056[0.03, 0.11]Small5–10 years2.67More than 10 years2.50SOL6: Library motivationLess than 5 years3.0025.4 < .0010.118[0.07, 0.19]Moderate5–10 years2.33More than 10 years2.63SOL7: Interactive activitiesLess than 5 years3.009.15.0100.036[0.004, 0.09]Small5–10 years2.92More than 10 years3.00SOL8: AI GuidelinesLess than 5 years2.764.34.1140.012[-0.01, 0.10]Small5–10 years2.92More than 10 years2.75SOL9: Plagiarism toolsLess than 5 years2.7615.0 < .0010.066[0.02, 0.19]Moderate5–10 years2.92More than 10 years3.00 Differences in perceptions of the solutions associated with using ChatGPT According to experience. Table 11Post-hoc pairwise comparisons (Dunn’s Test) for experience.VariableGroup 1Group 2z-statisticAdjusted pSignifSOL2 < 5 yrs > 10 yrs-3.33.003**5–10 yrs > 10 yrs-3.81 < .001***SOL3 < 5 yrs5–10 yrs-6.39 < .001****5–10 yrs > 10 yrs6.68 < .001****SOL5 < 5 yrs5–10 yrs-3.38.002** < 5 yrs > 10 yrs-3.19.004**SOL6 < 5 yrs5–10 yrs-2.71.020* < 5 yrs > 10 yrs-4.99 < .001****SOL75–10 yrs > 10 yrs-2.64.025*SOL9 < 5 yrs5–10 yrs3.88 < .001*** < 5 yrs > 10 yrs2.54.033*p-values are adjusted using the Bonferroni correction. Comparisons are shown for variables with significant overall differences (p < .05). Post-hoc pairwise comparisons (Dunn’s Test) for experience. p-values are adjusted using the Bonferroni correction. Comparisons are shown for variables with significant overall differences (p < .05). The strongest difference was observed for curriculum updating (SOL3), which showed a large effect size (ηH2 = 0.279). Dunn–Bonferroni post-hoc comparisons indicated significant differences between the less-than-5-years and 5–10-years groups, as well as between the 5 and 10-years and more-than-10-years groups. Support was highest among faculty with less than 10 years of experience, while faculty with more than 10 years reported lower support. This suggests that less experienced faculty may view curriculum renewal as a more urgent response to generative AI integration. Other differences were observed for in-person testing (SOL5), library-based work (SOL6), interactive activities (SOL7), and plagiarism detection tools (SOL9). Faculty with less than 5 years of experience showed stronger support for direct instructional safeguards, such as in-person assessments and library-based work. In contrast, faculty with more than 10 years of experience reported stronger support for plagiarism detection tools, suggesting a preference for technological monitoring rather than physical or location-based controls. At the same time, some solutions showed no significant differences across experience groups, including awareness initiatives (SOL1), permissible-use policies (SOL4), and AI guidelines (SOL8). This indicates a broader consensus that awareness and institutional guidance are necessary baseline responses regardless of professional experience. Findings suggest that professional experience influences how faculty prioritize AI-related interventions, with less experienced faculty favoring immediate instructional safeguards and more experienced faculty placing relatively higher emphasis on detection-oriented approaches. The Kruskal–Wallis H test results (Table 12) and subsequent post-hoc Dunn-Bonferroni pairwise comparisons (Table 13) reveal significant disparities in how various academic stakeholders perceive the challenges associated with implementing ChatGPT in higher education. Table 12Differences in perceptions of the challenges associated with using ChatGPT by years of experience.VariableDept. HeadFacultyDeanUniv. VPVice DeanHp-valueηH295% CIMagnitudeCH1: Biased info3.0002.3712.6002.3332.2868.93.0120.035[-0.00, 0.13]SmallCH2: Easier to cheat3.0003.0002.6002.3333.00029.8 < .0010.140[0.08, 0.21]ModerateCH3: Unequal evaluation3.0002.7302.6002.6673.00054.4 < .0010.263[0.18, 0.39]LargeCH4: Inconsistent values3.0002.5512.2002.6673.00012.0.0020.050[0.00, 0.12]SmallCH5: Fake citations2.4712.5512.4002.6673.00013.9 < .0010.060[0.01, 0.13]SmallCH6: Critical thinking3.0002.7302.6002.6673.00013.2.0010.056[0.03, 0.12]SmallCH7: Academic performance3.0003.0002.0002.6673.00015.6 < .0010.068[0.03, 0.13]ModerateCH8: Limit creativity3.0002.4612.6002.6672.7501.53.4650.000[-0.01, 0.04]SmallCH9: Data privacy2.5292.5512.4002.3333.0008.83.0120.034[0.00, 0.09]SmallCH10: Research skills2.0002.5512.2002.6673.0002.67.2630.003[-0.01, 0.06]SmallCH11: Human interaction3.0002.6292.4002.6672.7505.53.0620.017[-0.01, 0.11]Small Differences in perceptions of the challenges associated with using ChatGPT by years of experience. Table 13Significant pairwise comparisons (Post-hoc Dunn-Bonferroni).VariableComparison (Group A vs Group B)Statisticp.adjSignifCH1Dept. Head vs College Dean-3.99 < .001***Dept. Head vs Vice Dean-3.62.003**Faculty vs Dept. Head3.18.015*CH2Faculty vs Vice Dean-8.57 < .001****Dept. Head vs Vice Dean-6.22 < .001****College Dean vs Vice Dean-7.30 < .001****Univ. VP vs Vice Dean-5.06 < .001****Faculty vs Univ. VP-3.49.005**CH3College Dean vs Univ. VP-4.04 < .001***Dept. Head vs Univ. VP-3.31.009**College Dean vs Vice Dean-2.96.031*CH4Dept. Head vs Univ. VP-4.44 < .001****College Dean vs Univ. VP-5.42 < .001****Faculty vs Univ. VP-3.43.006**Faculty vs College Dean3.07.022*CH5Dept. Head vs College Dean3.40.007**College Dean vs Univ. VP-3.77.002**Faculty vs College Dean3.06.022*CH6College Dean vs Vice Dean-3.09.020*CH7Faculty vs Univ. VP-8.81 < .001****Dept. Head vs Univ. VP-5.79 < .001****College Dean vs Univ. VP-7.07 < .001****Faculty vs Vice Dean-3.63.003**College Dean vs Vice Dean-3.09.020*Univ. VP vs Vice Dean3.26.011*CH9College Dean vs Vice Dean-4.66 < .001****Faculty vs College Dean3.61.003**Dept. Head vs College Dean2.96.031*College Dean vs Univ. VP-3.80.001**CH10Dept. Head vs College Dean5.88 < .001****Faculty vs Dept. Head-4.41 <.001***College Dean vs Univ. VP-4.11 <.001***Dept. Head vs Vice Dean3.71.002**Faculty vs College Dean2.90.038*CH11Dept. Head vs Univ. VP-3.22.013* Significant pairwise comparisons (Post-hoc Dunn-Bonferroni). The most pronounced difference identified in this study relates to Unequal Evaluation (CH3), which yielded a large effect size (eta^2_H = 0.263, p < .001). Post-hoc analysis showed significant friction between College Deans and higher administration (Univ. VP and Vice Deans). Specifically, Vice Deans expressed significantly higher levels of concern compared to College Deans (z = -2.96, p = .031). This suggests that while mid-level leadership (Deans) may be more optimistic about AI integration, senior oversight roles remain cautious about the systemic threat AI poses to standardized assessment. Similarly, Easier to Cheat (CH2) showed a moderate effect size (eta^2_H = 0.140, p < .001). The pairwise comparisons indicate a "Polarization of the Frontline," where Vice Deans showed extreme statistical divergence from every other group, including Faculty (z = -8.57, p < .001) and University VPs (z = -5.06, p < .001). This highlights a critical perceptual gap; those responsible for academic policy (Vice Deans) view the ease of cheating as an existential threat to academic rigor far more acutely than those in general faculty or executive leadership positions. Significant differences were observed regarding Academic Performance (CH7) (p < .001). Post-hoc results show that University VPs hold views that differ significantly from Faculty (z = -8.81, p < .001) and College Deans (z = -7.07, p < .001). The means suggest that top-tier leadership is less convinced that AI will inherently lead to poor academic performance compared to faculty members who interact with student outputs daily. Furthermore, Inconsistent Values (CH4) and Fake Citations (CH5) revealed significant variance. For CH4, University VPs were the primary source of divergence, differing significantly from Faculty, Dept. Heads, and Deans. This indicates that while teaching-level staff are concerned with the immediate output quality and citation accuracy of AI, executive leadership may be more concerned with the broader alignment of AI with institutional values. Interestingly, Limit Creativity (CH8) and Research Skills (CH10) showed the highest levels of consensus across positions. CH8 yielded a non-significant p-value (p = .465) and a negligible effect size (eta^2_H = -0.002). This implies a universal academic sentiment—regardless of rank—that AI’s impact on creativity is a shared concern that transcends administrative hierarchy. The findings suggest that the “Academic Hierarchy” significantly colors the perception of AI risk. Vice Deans and Dept. Heads consistently reported higher concern for integrity-based issues (CH2, CH3, CH10), likely due to their direct responsibility for handling misconduct cases. Conversely, University VPs showed a statistical trend toward moderate or specialized concerns, suggesting a more macro-level, strategic view of the technology. For successful AI policy implementation, institutions must bridge the gap between the high-integrity anxieties of mid-level administration and the strategic optimism of executive leadership. The Kruskal–Wallis H test results (Table 14) and subsequent Dunn–Bonferroni post-hoc comparisons (Table 15) indicate statistically significant differences in how academic stakeholders prioritize solutions for managing ChatGPT integration. Table 14Differences in Perceptions of the solutions associated with using ChatGPT According to Position.VariableDept. HeadFacultyDeanUniv. VPVice DeanHp-valueηH2 = 95% CIMagnitudeSOL1: Awareness3.0002.9103.0002.6673.0005.45.0660.017[0.00, 0.08]SmallSOL2: Training faculty3.0003.0002.8002.6673.00019.1 < .0010.086[0.04, 0.15]ModerateSOL3: Curricula update3.0002.5512.8002.6673.00057.5 < .0010.279[0.15, 0.38]LargeSOL4: Permissible %2.4712.8203.0002.0002.7501.70.4270.000[-0.01, 0.04]SmallSOL5: In-person tests3.0002.8203.0001.6672.50013.1.0010.056[0.03, 0.09]SmallSOL6: Library motivation3.0002.5513.0001.6672.50025.4 < .0010.118[0.07, 0.19]ModerateSOL7: Interactive acts3.0003.0003.0002.6673.0009.15.0100.035[0.00, 0.08]SmallSOL8: Guidelines3.0002.8993.0002.6672.5004.34.1140.011[-0.01, 0.07]SmallSOL9: Plagiarism tools3.0002.8993.0002.6673.00015.0 < .0010.065[0.01, 0.21]Moderate Differences in Perceptions of the solutions associated with using ChatGPT According to Position. Table 15Significant pairwise comparisons (Post-hoc Dunn-Bonferroni).VariableComparison (Group A vs Group B)Statisticp.adjSignifSOL1Faculty vs Vice Dean-3.91 < .001***Dept. Head vs Vice Dean-3.88.001**College Dean vs Vice Dean-4.56 < .001****Univ. VP vs Vice Dean-4.77 < .001****SOL2Faculty vs Vice Dean-5.35 < .001****College Dean vs Vice Dean-4.56 < .001****Dept. Head vs Vice Dean-3.88.001**Faculty vs Univ. VP-3.88.001**College Dean vs Univ. VP-3.11.018*SOL3Faculty vs College Dean3.43.006**College Dean vs Vice Dean-2.83.047*SOL4Faculty vs Vice Dean-6.22 < .001****Univ. VP vs Vice Dean-6.51 < .001****Dept. Head vs Univ. VP4.15 < .001***College Dean vs Vice Dean-4.12 < .001***Faculty vs Dept. Head-3.60.003**SOL5Faculty vs Vice Dean-9.20 < .001****Dept. Head vs Vice Dean-7.44 < .001****College Dean vs Vice Dean-6.24 < .001****Univ. VP vs Vice Dean-9.13 < .001****College Dean vs Univ. VP2.83.046*SOL6Univ. VP vs Vice Dean-7.96 < .001****Faculty vs Vice Dean-6.29 < .001****Dept. Head vs Vice Dean-6.48 < .001****College Dean vs Vice Dean-5.33 < .001****Faculty vs Univ. VP3.20.014*SOL7Faculty vs Vice Dean-7.41 < .001****Dept. Head vs Vice Dean-5.38 < .001****College Dean vs Vice Dean-6.31 < .001****Univ. VP vs Vice Dean-6.60 < .001****SOL8College Dean vs Univ. VP3.46.006**Univ. VP vs Vice Dean-3.75.002**Faculty vs Vice Dean-2.93.034*Dept. Head vs Vice Dean-3.05.023*SOL9College Dean vs Vice Dean-4.44 < .001****Univ. VP vs Vice Dean-4.64 < .001****Faculty vs Vice Dean-3.63.003**Dept. Head vs Vice Dean-3.78.002** Significant pairwise comparisons (Post-hoc Dunn-Bonferroni). Curriculum updating (SOL3) emerged as the most prominent area of divergence, with a large effect size (ηH2 = 0.279, p < .001). Dunn–Bonferroni comparisons showed that Faculty Members reported significantly lower support than College Deans (z = 3.43, p = .006), while administrative groups (e.g., Department Heads and Vice Deans) demonstrated higher mean scores compared to Faculty Members. This pattern suggests that administrative leadership tends to view curricular reform as a necessary institutional response, whereas faculty members may express relatively lower support due to implementation demands. Training faculty (SOL2) also showed a moderate effect size (ηH2 = 0.086, p < .001). Post-hoc results indicated that University VPs differed significantly from Faculty Members (z = -3.88, p = .001) and College Deans (z = -3.11, p = .018), suggesting lower prioritization of immediate professional development at the executive level. Significant differences were also observed for in-person testing (SOL5) and library-based approaches (SOL6) (p < .001). In both cases, University VPs reported the lowest levels of support, in contrast to higher agreement among Department Heads and College Deans. Post-hoc comparisons confirmed significant differences between University VPs and Faculty Members for SOL6 (z = 3.20, p = ..014). These findings indicate variation between roles more closely involved in operational practices and those focused on broader institutional strategy. Post-hoc results further revealed that Vice Deans differed significantly from other groups across several solutions, including awareness initiatives (SOL1), interactive activities (SOL7), and plagiarism detection tools (SOL9) (p < . 001). For example, in SOL7, Vice Deans reported significantly higher support compared to Faculty Members (z = -7.41, p < .001) and University VPs (z = -6.60, p < .001). This suggests that mid-level administrative roles may report higher support for engagement-based pedagogical approaches. The findings indicate that academic position is associated with variation in preferred AI-related interventions. While there is broad support for pedagogical and detection-based solutions, differences remain regarding more traditional measures such as in-person testing. These patterns highlight the importance of aligning institutional strategies with the perspectives of different academic roles to support effective and sustainable AI integration. Qualitative analysis A thematic analysis approach, following Virginia Braun and Victoria Clarke (2006), was used to analyze responses to the open-ended questions. This approach was selected for its flexibility in identifying patterns across qualitative data and its suitability for examining faculty perceptions in applied educational contexts. Participants were asked open-ended questions such as: “What are the main challenges associated with using ChatGPT in higher education?” and “What solutions do you suggest to address these challenges?”. The analysis followed a structured process. Responses were reviewed multiple times to ensure familiarity with the data. Initial codes were generated to capture recurring ideas and were then grouped into broader themes. These themes were refined, defined, and named based on their coherence and relevance across responses. Coding was conducted by the researchers and iteratively reviewed to enhance consistency and reduce subjective bias. NVivo 12 was used to support coding and data organization, enabling systematic classification and comparison across themes. Although not all participants provided extensive qualitative responses, the available data offered meaningful insights into recurring patterns that complement the quantitative findings. Table 16 summarizes the main themes identified from the qualitative analysis. The following sections discuss each theme in detail with supporting participant quotations. Table 16Summary of themes identified from the qualitative analysis.CategoryThemeBrief descriptionChallengesWeakness in writingConcerns that reliance on ChatGPT reduces students’ independent writing and research skillsChallengesCheating and plagiarismConcerns about academic misconduct and difficulties in verifying the originality of students’ workChallengesWeakness in cognitive and communication skillsConcerns regarding reduced critical thinking, communication, and independent learningChallengesAcademic integrityBroader ethical concerns related to the responsible use of generative AI in higher educationSolutionsProfessional developmentContinuous training and AI literacy initiatives to support faculty members in the effective and responsible use of AISolutionsInstitutional regulationsDevelopment of clear university policies and governance frameworks to regulate the responsible use of AI Summary of themes identified from the qualitative analysis. Qualitative findings: challenges Table 17 presents the most frequently reported themes related to the perceived educational impact of AI use among students. The most frequently reported theme was weakness in writing skills (16%). Participants expressed concern that reliance on AI tools may reduce students’ engagement in independent writing. One respondent noted, “Everything has become easy for students; writing research and assignments now involves artificial intelligence,” while another stated, “They use AI directly to complete their tasks or write papers.” Table 17Emerging Themes on Challenges of ChatGPT Use in Saudi Universities.ThemesN%Weakness in writing (reports, research)3216Cheating and plagiarism2412Weakness in some skills (thinking, communication, and human interaction)189Academic integrity168 Emerging Themes on Challenges of ChatGPT Use in Saudi Universities. This theme helps explain the quantitative findings, particularly the relatively higher concern regarding academic performance and skill development. The second theme, cheating and plagiarism (12%), may reflect concerns about academic misconduct and difficulties in verifying authorship. Participants highlighted challenges in distinguishing original work from AI-generated content. One respondent stated, “AI programs have increased, and I believe they have led to a significant rise in academic cheating,” while another noted that “it is becoming harder to verify whether the work submitted is the student’s own effort.” These qualitative insights provide explanatory depth for the quantitative results related to cheating and unequal evaluation. The theme of weakness in cognitive and communication skills (9%) captures concerns about the impact of AI on critical thinking, communication, and interaction. Participants indicated that reliance on AI-generated responses may limit opportunities for independent thinking. One respondent noted that “reliance on ready-made answers reduces students’ ability to think critically,” while another stated that “students may lose communication and discussion skills over time.” This theme supports quantitative findings related to critical thinking and reduced human interaction. Finally, academic integrity (8%) emerged as a key concern, emphasizing ethical considerations and responsible AI use. One respondent noted, “AI use raises questions about honesty and responsibility.” While related, “cheating and plagiarism” refer to specific student practices, whereas “academic integrity” reflects broader ethical concerns surrounding the use of AI. This theme reinforces the emphasis on academic integrity observed in the quantitative analysis. Qualitative findings: solutions The qualitative findings also identified key themes related to proposed solutions for addressing ChatGPT-related challenges. Table 18 presents the most prominent themes related to institutional and educational strategies for AI integration. The most prominent theme was professional development (20%). Participants emphasized the importance of training faculty to engage effectively with AI tools. One respondent noted that “continuous training is needed to help faculty use AI responsibly,” while another highlighted that “professional development programs can improve understanding of AI in teaching.” Table 18Emerging themes on solutions for ChatGPT Use in Saudi universities.ThemesN%Professional development4120Developing regulations in universities2814 Emerging themes on solutions for ChatGPT Use in Saudi universities. This theme provides qualitative support for the quantitative findings, where training and awareness received strong levels of agreement. The second theme focused on institutional regulations (14%). Participants stressed the need for clear policies to guide AI use and prevent misuse. One respondent stated that “universities should establish clear rules for AI use in assignments and research,” while another noted that “regulations are necessary to control misuse of AI tools.” These findings explain the strong quantitative support for institutional guidelines and governance mechanisms. Integration with quantitative findings The qualitative findings provide explanatory insight into the quantitative patterns observed across academic rank and experience. For example, the strong emphasis on academic integrity and cheating helps explain why faculty reported elevated concern in the quantitative analysis, particularly in relation to cheating and unequal evaluation. Similarly, concerns about writing skills and critical thinking provide context for quantitative findings related to academic performance and reduced engagement. In addition, the emphasis on professional development and institutional regulations explains the strong support for training and policy-based solutions identified in the quantitative results. These patterns can be interpreted in light of differences in professional roles and responsibilities within higher education, where those more directly engaged in teaching and assessment may be more sensitive to practical challenges, while administrative roles may prioritize broader institutional strategies. Discussion The findings indicate that faculty perceptions of ChatGPT are primarily shaped by concerns about academic integrity, assessment validity, and the implications of generative AI for teaching practice. These concerns, however, are not uniform across faculty members. Instead, they vary according to academic rank, years of professional experience, and administrative responsibilities. Such variation appears to reflect differences in academic duties, institutional expectations, and the nature of faculty engagement with teaching, assessment, and educational governance. Within the Saudi higher education context, these differences may also be influenced by varying levels of institutional readiness and the ongoing development of policies governing the educational use of artificial intelligence. Concerns related to cheating, plagiarism, and unequal evaluation emerged as the most prominent challenges. This finding is consistent with previous research highlighting the impact of generative AI on assessment practices and the increasing difficulty of verifying authorship20,32,63. The present findings suggest that these concerns are not merely theoretical but are embedded in everyday assessment practices. This interpretation is further supported by evidence indicating that students may normalize or conceal their use of AI tools. Collectively, these findings suggest that maintaining academic integrity requires redesigning assessment practices rather than relying solely on restrictive measures. Differences across academic rank indicate that faculty perceptions are influenced by role-related expectations and academic responsibilities. Associate Professors reported the highest levels of concern, particularly regarding assessment credibility and student performance. This pattern is broadly consistent with previous studies showing that faculty responses to generative AI differ across academic roles7,36. One possible interpretation is that Associate Professors often balance intensive teaching, student assessment, curriculum development, quality assurance, and expanding institutional service responsibilities. These overlapping demands may increase their sensitivity to risks that directly influence educational quality and assessment credibility. This interpretation aligns with Role Theory, which proposes that individuals interpret workplace challenges according to the expectations and responsibilities associated with their professional roles64. It is further supported by research demonstrating that academic staff simultaneously perform multiple academic and organizational functions, leading to distinct perspectives across academic positions65. Assistant Professors also reported relatively high levels of concern, possibly reflecting their closer involvement in day-to-day teaching, assessment, and direct interaction with students. From the perspective of Diffusion of Innovation Theory, faculty who engage more directly with classroom practice are often among the first to experience both the opportunities and challenges associated with emerging technologies, while their responses remain shaped by the surrounding institutional environment and academic culture66,67. In contrast, Professors expressed comparatively lower levels of concern, which may reflect greater involvement in research leadership, strategic planning, and institutional decision-making. Consequently, they may be more likely to view generative AI from a broader institutional perspective than primarily as a classroom challenge. Professional experience also appears to shape faculty perceptions. Less experienced academics expressed stronger support for training initiatives and changes in assessment practices. This finding is consistent with research suggesting that early-career faculty tend to be more receptive to technological innovation when appropriate institutional support is available6. By contrast, more experienced faculty reported greater concern regarding reduced human interaction. This observation is consistent with previous studies emphasizing the importance of interpersonal relationships in teaching and learning when evaluating educational technologies26. It may also reflect stronger attachment to established teaching practices developed through years of classroom experience, leading experienced faculty to place greater value on direct interaction with students and colleagues. Administrative responsibilities further influenced how faculty interpreted both challenges and proposed solutions. Department Heads and Vice Deans reported higher levels of concern and stronger support for policy-oriented responses. This pattern may reflect their responsibility for maintaining academic standards, overseeing assessment quality, managing institutional risk, and ensuring compliance with university policies. Institutional Theory provides a useful explanation for this finding by suggesting that academic leaders play a central role in protecting institutional legitimacy through governance structures, formal regulations, and organizational policies68. Recent studies on generative AI governance similarly emphasize that effective institutional responses depend on leadership, clear regulations, ethical guidance, and coordinated governance frameworks rather than relying exclusively on individual faculty practices69. Organizational research further indicates that faculty develop different institutional perspectives according to their multiple academic identities and organizational responsibilities, leading academic leaders to interpret technological change from a broader institutional viewpoint65. Similar patterns have been reported in recent research7,36. In contrast, faculty without administrative responsibilities tended to emphasize more immediate classroom responses, including training, awareness, and practical guidance for teaching and assessment. The quantitative and qualitative findings converged around several common priorities. Professional development, clear institutional regulations, and assessment redesign consistently emerged as the most important responses. Differences between the two datasets are likely attributable to the fact that the qualitative analysis reflects a smaller group of participants, whereas the quantitative findings represent the full study sample. Nevertheless, the qualitative findings—particularly those concerning writing skills, cheating, and academic integrity—provide valuable context for understanding the quantitative results by illustrating how these challenges are experienced in everyday educational practice. Together, the two strands of evidence indicate that faculty concerns arise from practical teaching experiences rather than abstract perceptions of technology. The present study contributes to the growing literature on generative AI in higher education by demonstrating that faculty perceptions are shaped through the interaction of professional roles, teaching responsibilities, institutional expectations, and organizational context. Rather than simply documenting differences across faculty groups, the findings explain why these differences emerge. Faculty responses appear to reflect variations in role expectations, institutional responsibilities, and engagement with technological innovation. Unlike previous studies that primarily describe faculty attitudes toward generative AI, the present study provides a theoretically grounded explanation of how institutional position, academic role, and organizational responsibilities influence these perceptions within higher education. The findings therefore suggest that concerns surrounding ChatGPT are shaped not only by the technology itself but also by the institutional environments through which faculty experience its educational implications. They further indicate that issues relating to academic integrity, assessment, and teaching should be understood as interconnected components of broader institutional change rather than as isolated technological challenges. The findings have important practical implications for higher education institutions. They highlight the need to redesign assessment practices, strengthen mechanisms that support academic integrity, establish clear institutional policies governing the educational use of generative AI, and provide professional development that reflects the differing responsibilities of academic and administrative staff. These recommendations are particularly relevant for Saudi higher education institutions, where the rapid integration of artificial intelligence requires governance frameworks capable of balancing technological innovation with academic integrity, educational quality, and institutional trust. Limitations This study was conducted within Saudi universities, which may limit the generalizability of the findings to other higher education contexts. The cross-sectional design captures faculty perceptions at a single point in time and does not account for potential changes over time. The use of a 3-point Likert scale may have limited response variability and sensitivity, potentially constraining the granularity of the analysis despite the use of non-parametric methods. In addition, the qualitative findings are based on a subset of participants who responded to open-ended questions, which may not fully represent the perspectives of the entire sample. The study also did not control for potential confounding variables such as academic discipline and gender. Therefore, the findings should be interpreted with caution, as they may be influenced by these methodological and contextual factors. Practical implications The present findings support the development of clear institutional policies governing the educational use of generative AI and assessment practices that safeguard academic integrity. They also highlight the value of professional development to support faculty in AI-aware teaching and assessment, together with initiatives that promote AI literacy and the responsible use of generative AI among students. Conclusion This study examined faculty perceptions of the challenges and responses associated with the use of ChatGPT in Saudi higher education. The findings indicate that concerns are primarily centered on academic integrity and assessment practices, rather than on the technology itself. Faculty responses suggest that the main difficulty lies in maintaining credible evaluation in AI-mediated environments. Differences across academic rank, experience, and administrative roles highlight those perceptions of AI are shaped by professional responsibilities. Faculty who are more directly engaged in teaching and assessment tend to express stronger concern, while those in strategic roles emphasize institutional responses. These patterns suggest that AI-related challenges are experienced unevenly across the academic system. The study contributes to the literature by showing that concerns about generative AI are interconnected and closely linked to assessment design. Rather than treating academic integrity, learning, and institutional policy as separate issues, the findings indicate that they operate together within everyday teaching practice. This study underscores the need to align assessment practices, pedagogical approaches, and institutional policies to support the effective integration of generative AI in higher education. Strategies that account for differences in faculty roles and experience are likely to be more effective than uniform approaches. This suggests that the integration of generative AI in higher education should be understood as a socio-institutional process rather than a purely technological shift, requiring context-sensitive and role-aware implementation strategies. This includes supporting faculty—particularly those involved in teaching and assessment—through targeted training focused on AI-aware assessment design and academic integrity practices. These implications operate across multiple levels, including policy-level guidance, institutional practices, and individual faculty engagement. Future research should explore how faculty adapt assessment practices in response to generative AI over time and examine how disciplinary context, academic roles, and institutional settings shape these adaptations. Author contributions Salem Alalwani. conceptualized the study, conducted the literature review, performed the analysis, and wrote the manuscript. The author also reviewed and approved the final version of the manuscript. Funding This research received no external funding. Data availability The dataset generated and analyzed during the current study is included in the supplementary materials submitted with the manuscript. Declarations Ethics approval Ethical approval for this study was obtained from the Local Research Ethics Committee for Research Involving Living Creatures at the University of Hafr Al-Batin. The research protocol, survey instrument, and data collection procedures were reviewed and approved in accordance with institutional research regulations and internationally recognized ethical standards for research involving human participants. Approval was granted under Approval No. HPO-05-FT-25/18 in June 2024, prior to the commencement of data collection. The scope of the approval covered participant recruitment, data collection procedures, confidentiality safeguards, and data management practices. The study did not involve any clinical experimentation. Informed consent Informed consent was obtained from all participants prior to their participation in the study. Participants were fully informed about the purpose of the research, the voluntary nature of participation, the use of anonymized data for research purposes, and their right to withdraw from the study at any time without any consequences. Confidentiality and anonymity were assured. Informed consent was obtained electronically at the time of data collection, which was conducted between July 2024 and February 2025. AI use disclosure statement ChatGPT was used solely to improve language clarity. All scientific content and interpretations are the author’s own. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The dataset generated and analyzed during the current study is included in the supplementary materials submitted with the manuscript.