Abstract
The rapid growth of artificial intelligence (AI), particularly generative AI technologies, is transforming higher education and redefining the competencies required for the future workforce. In this context, transversal skills such as critical thinking, collaboration, digital competence, and ethical reasoning are increasingly recognized as essential for engaging effectively with AI-driven environments. Despite growing scholarly attention to AI in education, the broader research landscape connecting transversal skills and AI in higher education remains insufficiently mapped. This study addresses this gap through a bibliometric analysis of 1,487 Scopus-indexed publications published between 2020 and 2025. Using Bibliometrix and VOSviewer, the analysis examines publication trends, collaboration patterns, leading countries and institutions, and the conceptual structure of the field. The findings reveal a rapidly expanding research landscape, particularly following the emergence of generative AI tools. AI literacy, digital competence, and critical thinking emerge as the most prominent competencies within the literature. The analysis also identifies five major research themes centered on curriculum transformation, pedagogical innovation, AI literacy and digital competence, educator capacity building, and professional readiness. Emerging topics such as metacognition, ethical reasoning, and workforce readiness suggest a growing shift toward human-centered competencies that support effective human–AI collaboration and prepare graduates for AI-driven future work environments.
Keywords
Introduction
The pervasive expansion of AI is fundamentally reshaping the educational landscape, global labor markets, and professional practices. In this context, transversal skills have gained attention as an important aspect of higher education, preparing graduates for the AI-driven future. Although transversal skills are not a new concept in educational contexts, which have long been recognized as an essential set of skills that make graduates more employable (Wilkinson, 2024), the emergence of generative AI tools has heightened the need for transversal skills more than ever, as AI technologies tend to transform the future of work in many sectors significantly (Zahn et al., 2025). As AI systems increasingly automate routine cognitive and technical tasks, questions have emerged about which human capabilities remain essential. Skills that enable individuals to adapt, collaborate, and think critically in AI-driven environments are increasingly recognized as part of a growing partnership between humans and intelligent systems (Brynjolfsson & McAfee, 2014; Kong et al., 2025; World Economic Forum, 2023). Specifically, cognitive skills such as critical thinking now play a central role in enabling learners to interpret algorithmic outputs, question the decisions generated by automated systems, and make informed judgments in AI-mediated environments (Mayer et al., 2025). Many researchers have reported this shift, often called the “Human-AI partnership,” in which people and AI work together, combining human skills and AI capabilities (Brynjolfsson & McAfee, 2014; Jarrahi, 2018). Virgilio et al. (2024) emphasize that the “human touch,” including interpersonal judgment, cultural sensitivity, and nuanced communication, remains difficult to replicate algorithmically.
Labor market studies echo this trend, showing that workers with strong transversal skills are less vulnerable to automation and better positioned to complement AI and machine-learning systems (Jetha et al., 2021). These shifts carry direct consequences for how we understand graduate employability. The World Economic Forum (2023) identified analytical thinking, creative thinking, resilience, flexibility, and curiosity among the top skills employers will prioritize through 2027, all of which fall squarely within the domain of transversal skills. The traditional model of employability, built around discipline-specific knowledge and technical qualifications, is gradually giving way to a more dynamic understanding in which adaptability, interdisciplinary collaboration, and continuous learning are as important as subject expertise (Succi & Canovi, 2020; Tomlinson, 2017). Across sectors, employers increasingly report that graduates who communicate effectively, work well in teams, reason ethically, and regulate their own learning are better prepared to navigate ambiguity and contribute meaningfully to roles that demand human judgment alongside AI systems (Bridgstock, 2009; Jackson, 2016).
Higher education institutions increasingly frame graduate employability as a combination of AI literacy and transversal skills, emphasizing the importance of competencies that enable students to interpret AI outputs, collaborate across disciplines, and engage critically with data-driven technologies (Chicioreanu & Oproiu, 2025; Dumitru & Halpern, 2023; Michalon & Camacho-Zúñiga, 2023; Walter, 2024). Consequently, institutions are exploring pedagogical approaches that foster critical inquiry, collaborative problem-solving, and digital literacy within AI-mediated learning environments (Chiu et al., 2022). In this context, transversal skills are no longer viewed merely as desirable graduate attributes but as essential competencies for meaningful participation in an AI-driven workforce. Understanding how research conceptualizes these skills in higher education has therefore become increasingly important.
Yet despite this growing scholarly interest, the research landscape surrounding transversal skills in the context of generative AI in higher education remains fragmented and insufficiently mapped (Kang et al., 2025; Kavitha et al., 2024; Nain et al., 2025). A bibliometric approach is particularly well suited to this purpose because it enables a systematic, data-driven examination of how a rapidly expanding field evolves by tracking publication trends, identifying dominant themes, and revealing collaboration structures that individual studies may not capture (Kavitha et al., 2024). Accordingly, this study aims to map the evolving research landscape of transversal skills in higher education in the age of AI. Using bibliometric analysis of Scopus-indexed publications published between 2020 and 2025, the study examines publication trends, collaboration structures, key competencies, and emerging thematic areas within the field. Through this analysis, the study identifies which aspects of the transversal skills research landscape are well established, which are emerging, and which remain underexplored, thereby offering important implications for future research prioritization, curriculum development, and educator capacity building in higher education.
Literature Review
Defining and Conceptualizing Transversal Skills
Transversal skills represent a broad and multifaceted category of competencies that transcend specific occupational domains, disciplines, or tasks. They are often referred to interchangeably as “soft skills,” “generic skills,” “transferable skills,” or “21st-century skills” (Deroncele-Acosta et al., 2025; Marle et al., 2022; OECD, 2019; Thornhill-Miller et al., 2023; UNESCO, 2014; Wilkinson, 2024; Zahn et al., 2025). As UNESCO (2014) explains, transversal skills are widely understood as skills that are “reusable and transferable from one field to another and not related to a specific job, task, discipline, or occupation.” These skills encompass a combination of knowledge, attitudes, dispositions, and behavioral capacities that support effective collaboration, adaptability, communication, ethical reasoning, and problem-solving. They enable individuals to navigate complex and rapidly changing social, technological, and economic environments (Deroncele-Acosta et al., 2025; OECD, 2019; Weber et al., 2025; Zhan et al., 2024).
In higher education, transversal skills are increasingly positioned as foundational competencies that complement disciplinary expertise and enhance graduate employability. However, the conceptual boundaries of transversal skills remain somewhat fluid, reflecting both the evolving nature of work and the diverse disciplinary perspectives from which these competencies are studied (UNICEF, 2019; Weber et al., 2025). To provide a coherent analytical lens for this bibliometric analysis, several internationally recognized competency frameworks were used to define and guide the conceptual scope of transversal skills examined in this study.
First, UNICEF’s (2019) Global Framework on Transferable Skills conceptualizes transferable skills within a holistic learning framework based on four dimensions of learning: cognitive, instrumental, individual, and social. From these dimensions, the framework identifies twelve transferable skills, including skills such as critical thinking, problem-solving, self-management, and communication, which support individuals’ development across domains such as learning, employability, personal empowerment, and active citizenship. Second, the International Labor Organization (ILO, 2021) framework for core skills for life and work in the twenty-first century groups transferable skills into four categories: social and emotional skills, cognitive and metacognitive skills, basic digital skills, and foundational skills for green jobs. More recently, Weber et al. (2025), drawing on psychological research and educational practice, proposed a framework consisting of cognitive skills, citizenship, well-being, and social-emotional skills as core transversal skill categories. These categories include skills such as critical thinking, creativity, and digital literacy (cognitive skills); ethical awareness and civic responsibility (citizenship); self-regulation, motivation, and personal well-being (well-being); and interpersonal skills such as communication, collaboration, leadership, and conflict resolution (social-emotional skills). Collectively, these frameworks provide the conceptual lens through which transversal skills are interpreted and analyzed in this study.
In the context of rapid technological change, research on transversal skills in higher education can be understood as operating at the intersection of three key dimensions: technological transformation, educational practice, and workforce preparedness. The rapid development of generative AI technologies is redefining how knowledge is produced, evaluated, and applied, prompting higher education institutions to reconsider the skills students need to thrive in AI-mediated environments (Yurdal, 2025). Consequently, examining how the academic literature conceptualizes transversal skills in relation to artificial intelligence provides important insights into how higher education research is responding to the evolving demands of AI-driven work environments.
Research Gap
In labor markets characterized by rapid technological change, having disciplinary knowledge alone is no longer sufficient. This is because technical expertise can gradually become outdated, whereas transversal skills support mobility across different roles and sectors (Bridgstock, 2009). Studies suggest that as AI reshapes work, the concept of graduate employability is undergoing a significant shift. Employers increasingly seek graduates who can adapt, collaborate, and innovate in dynamic environments (Sozon et al., 2026; Cingillioglu & Schoettner, 2025; Ragusa et al., 2025; Zhan et al., 2024; Yong & Ling, 2022). In AI-driven workplaces, transversal skills are especially important because they enable professionals to interpret complex outputs, collaborate across disciplines, and navigate ethical challenges associated with algorithmic systems. In this context, preparing graduates for the future workforce and providing them with opportunities to foster transversal skills remains of paramount importance. Examining the literature, the intersection of transversal skills and AI in higher education has attracted growing scholarly interest.
Several recent review studies have contributed to the understanding of transversal skills in the context of AI and higher education, each from distinct but complementary perspectives. Wilkinson (2024) engaged in a thematic discussion in generic skills in relation to large language models, but the researcher has not followed a bibliometric analysis approach. Villegas (2024) and Amarathunga et al. (2024) examined soft skills for employability but did not specifically focus on the AI dimension. Meanwhile, existing bibliometric studies on generative AI in higher education (e.g., Batubara et al., 2024; Henukh et al., 2025; Kavitha et al., 2024; Mittal et al., 2026) map the broader field of AI in higher education but do not specifically focus on transversal skills.
The most recent study related to transversal skills in higher education is Deroncele-Acosta et al. (2025). They conducted a systematic review on AI and transversal skills in higher education. Their study identified a range of transversal skills and strategic processes for the sustainable integration of generative AI in higher education. The analysis primarily relied on narrative synthesis and was supported by bibliometric techniques to map the broader structural and thematic patterns of the field. However, the scope of the study was limited to the 2023–2025 period. Similarly, Idrus et al. (2025) applied a bibliometric approach to map generative AI research in education but focused exclusively on critical thinking as a single competency rather than examining the broader range of transversal skills required for graduate employability in the AI era. Further, Zahn et al. (2025) mapped soft skills and future research directions in higher education; however, their analysis covered the period from 1982 to 2022 and did not specifically examine the emergence of generative AI.
Despite growing interest in AI and graduate skills, the broader research landscape on transversal skills in the context of generative AI in higher education remains insufficiently mapped. In particular, there is a lack of comprehensive bibliometric analyses examining how scholarship has evolved during the recent surge of generative AI technologies. As Kavitha et al. (2024) argue, navigating the broad and rapidly evolving landscape of AI in education necessitates a more systematic and comprehensive approach to grasp its trends and emerging areas. Kang et al. (2025) similarly contend that bibliometric analysis offers a powerful tool for mapping research hotspots and pinpointing key areas for development, while Nain et al. (2025) highlight its capacity for multidisciplinary citation tracking across a growing field. In the specific context of generative AI which evolves rapidly, Polat et al. (2024) demonstrated how bibliometric methods can surface exponential growth patterns and dominant thematic clusters that conventional reviews cannot capture at scale. Thus, it is essential to examine the visibility of these publications and their practical implications for the advancement of the field, as doing so provides an evidence base for informed decision-making by researchers, educators, and policymakers, making this analysis both timely and methodologically necessary.
To address this gap, this study employs a bibliometric approach to analyze research on transversal skills in higher education in the age of AI between 2020 and 2025. By examining publication trends, collaboration patterns, dominant competencies, and thematic developments, the study provides a comprehensive overview of how scholarship on transversal skills in higher education is evolving in response to the rapid expansion of generative AI technologies.
To guide this analysis, the study addresses the following research questions:
By providing a comprehensive bibliometric mapping of transversal skills research in the age of generative AI, this study contributes to understanding how higher education scholarship is responding to the changing skill demands of AI-mediated work environments.
Methodology
Analytical Framework
This study adopts a bibliometric research design to systematically map research on transversal skills in higher education in the age of AI. The analysis follows established methodological guidelines for science mapping and bibliometric research (Aria & Cuccurullo, 2017; Donthu et al., 2021; Zupic & Čater, 2015).
To ensure comprehensive coverage, the Scopus database was used as the primary source of records. Managed by Elsevier, Scopus is one of the largest abstract and citation databases and provides extensive coverage across multiple disciplines (Zhan et al., 2024). It also supports high-quality metadata exports compatible with bibliometric tools such as Bibliometrix and VOSviewer, making it particularly suitable for mapping research structures and trends (Donthu et al., 2021). Consequently, Scopus has been widely used in bibliometric studies due to its broad disciplinary coverage and well-structured citation metadata (Mittal et al., 2026).
Systematic Search
The search was conducted on 15 February 2026. Search keywords related to (1) AI technologies, (2) transversal skills, and (3) higher education were identified through a preliminary literature review. The retrieval search string was developed using keywords and Boolean operators. To ensure the relevance and accuracy of the search results, the search query was iteratively refined by reviewing key publications in the field and adjusting keywords to capture commonly used variations in the literature. The following search string was applied to titles, abstracts, and keywords in Scopus using the TITLE-ABS-KEY field tag:
TITLE-ABS-KEY ((“artificial intelligence” OR “AI” OR “deep learning” OR “generative AI” OR “generative artificial intelligence” OR “ChatGPT” OR “large language model*” OR “LLM*” OR “machine learning”)) AND (“transferable skill*” OR “transversal skill*” OR “generic skill*” OR “21st century skill*” OR “employability skill*” OR “future skill*” OR “competenc*”) AND (“higher education” OR “universit*” OR “graduate education” OR “postgraduate” OR “tertiary education” OR “college education”))
The search query consisted of three conceptual blocks. The first block targeted artificial intelligence–related terminology, such as “artificial intelligence,” “AI,” “deep learning,” “generative AI,” and “ChatGPT,” capturing both the broader AI concept and technologies associated with the current AI landscape in education. The second block focused on terminology related to transversal skills. Although this study adopts “transversal skills” as the primary term, the literature uses several related expressions, including “transferable skills,” “generic skills,” “21st-century skills,” “employability skills,” “future skills,” and “competencies.” These synonymous terms were included to ensure that relevant studies were not excluded due to disciplinary variations in terminology. The third block restricted the search to the higher education context by including terms such as “higher education,” “university,” “graduate education,” “postgraduate,” “tertiary education,” and “college education.”
To ensure transparency and reproducibility in the study selection process, this research followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) framework for screening and selecting documents. The initial search returned 3240 records. After applying the eligibility criteria, 1,487 documents were retained for the final bibliometric analysis. The PRISMA flow diagram illustrating the document screening process is presented in Figure 1. PRISMA Flow Diagram for The Systematic Search and Selection Process
Eligibility Criteria
A set of explicit inclusion and exclusion criteria was used to ensure the relevance of the articles as follows: (1) (2) (3) (4)
Bibliometric Analysis Tools
This study employed two widely used bibliometric tools: (1) Bibliometrix and (2) VOSviewer to process and analyze the retrieved dataset. Bibliometrix, an open-source R package for comprehensive science mapping analysis (Aria & Cuccurullo, 2017), was used for data processing and descriptive analysis. Its web interface, Biblioshiny (version 5.0), was used to generate key bibliometric indicators, including annual publication trends, most productive authors, journals, institutions, and countries. VOSviewer was used to construct and visualize bibliometric networks, including keyword co-occurrence maps and thematic clusters (van Eck & Waltman, 2010). These visualizations enabled the identification of dominant themes, emerging research trends, conceptual clusters, and the most frequently discussed transversal skills within the literature.
Data Analysis
Dataset Overview and Descriptive Statistics
The dataset comprises 1,487 Scopus-indexed documents published between 2020 and 2025. These documents were authored by 5,597 authors and cite 11,789 references. In total, 4,047 unique author keywords were identified, reflecting the terminological diversity and multidisciplinary nature of research on transversal skills in higher education in the age of AI.
The field shows rapid growth (annual growth rate = 94.95%), with the strongest increase occurring after 2022 alongside the expansion of generative AI. The average document age is 1.8 years, indicating a highly recent literature. Documents receive an average of 14.14 citations, suggesting active scholarly engagement. The average number of authors per document is 4.14, indicating a preference for team-based research, although 181 documents (12.2%) are single-authored. International co-authorship is 23.6%, suggesting moderate cross-border collaboration and scope for deeper international research networks.
Together, these descriptive statistics highlight the rapid growth and collaborative nature of research on transversal skills in the age of AI.
Annual Scientific Production Trends
To answer RQ1, Figure 2 shows annual publication trends on transversal skills in the age of AI within higher education research from 2020 to 2025. The trajectory indicates rapid growth with three phases. From 2020 to 2022, annual output remained below 100 publications. In 2023, production rose to 133 publications, up from 31 in 2020, marking a clear inflection point. This increase aligns with the public release of ChatGPT in November 2022 and the subsequent expansion of generative AI, which intensified research attention to education, assessment, and workforce-relevant transversal skills. The steepest increase occurred in 2024–2025: publications more than doubled to approximately 310 in 2024 and rose further to over 850 in 2025, the highest annual output in the dataset. Overall, the annual growth rate (94.95%) indicates that this topic has become a rapidly expanding area of higher education research. Annual Scientific Production Trends (2020–2025)
Taken together, these publication trends indicate a rapidly expanding research area, particularly after the emergence of generative AI. The following sections examine the sources, authors, and collaboration structures that are shaping this growing body of scholarship.
Most Productive Sources
Most Productive Sources by Publication Volume (2020–2025)
The distribution of sources reflects the interdisciplinary scope of research on AI and transversal skills, extending beyond traditional education journals to include medical education, technology-focused, and multidisciplinary outlets. The prominence of journals such as BMC Medical Education and Journal of Surgical Education suggests that AI-related workforce transformation is particularly active in health professions education, where developments in diagnostics, simulation, and clinical decision-making are accelerating the demand for transversal competencies. This pattern indicates that the influence and adoption of AI are not uniform across disciplines; rather, certain fields, particularly medicine, appear to engage more intensively with AI-driven educational and workforce changes due to the rapid technological evolution of professional practice.
Most Influential Sources Across Disciplines in AI and Transversal Skills Research
Note. NP = Number of Publications; TC = Total Citations; ACPP = Average Citations per Publication; SNIP = Source Normalized Impact per Paper. SNIP values are retrieved from the Scopus database (2024).
Computers and Education: Artificial Intelligence records the highest total citation count (1,714 citations from 21 publications), indicating a prominent role in scholarly discussions on AI and transversal skills. In contrast, the International Journal of Educational Technology in Higher Education achieves the highest citation intensity, accumulating 1,547 citations from only eight publications and the highest ACPP value (193.38). This pattern suggests that a relatively small number of highly influential studies have significantly shaped research visibility within the field. A similar trend is visible in Academic Medicine, which generates 286 citations from only 10 publications (ACPP = 28.60), reflecting the strong visibility of AI-related transversal skills research within medical education scholarship.
In comparison, journals such as BMC Medical Education, Education Sciences, and Frontiers in Education contribute substantial publication volume but demonstrate more moderate citation influence per article. These patterns suggest that visibility within the field is shaped not only by publication volume but also by the presence of conceptual and empirical contributions that attract citation attention. The presence of journals such as IEEE Access and PLOS ONE further highlights the topic’s diffusion beyond education-focused outlets into engineering, computer science, and multidisciplinary research domains.
Overall, the findings suggest that AI and transversal skills research remain an emergent and intellectually dispersed field characterized by concentrated citation impact and cross-disciplinary knowledge exchange. Rather than being dominated by a single disciplinary tradition, the field is being shaped through contributions from multiple disciplinary communities, reflecting the broad educational and societal relevance of AI-related transversal competencies.
While journal-level analysis highlights where research is published, the next section, examining author productivity and collaboration patterns provides further insight into the intellectual structure of the field.
Author Contribution and Collaboration Patterns
Most Productive Authors
Top 15 Most Productive Authors With Impact Metrics (2020–2025)
Note. NP = number of publications; TC = total citations within the dataset; ACPP = average citations per publication; h-index = number of publications (h) receiving at least h citations within this corpus; g-index = largest number (g) such that the top g publications received at least g2 citations collectively; m-index = h-index divided by the number of years since the author’s first publication in the dataset; PY_start = year of first publication in the corpus. All citation indicators were calculated locally from the dataset of 1,487 documents.
Among the authors, T.K.F. Chiu and Y. Wang are the most productive, each contributing nine publications. Chiu records the highest overall citation impact (TC = 879), the highest h-index (6), and a high ACPP value (97.67), reflecting both sustained research activity and strong citation visibility. In contrast, Wang demonstrates similar publication output but lower citation influence per article (ACPP = 14.89). These differences show that publication volume alone does not necessarily correspond to citation impact.
The value of ACPP becomes more evident when examining authors with comparatively fewer publications. B.L. Moorhouse records the highest ACPP in the table (134.20) from five publications and 671 total citations, indicating substantial per-paper influence despite lower overall productivity. Similarly, J. Lee (ACPP = 72.40), G.-J. Hwang (ACPP = 55.50), and Del Maestro (ACPP = 42.67) demonstrate strong citation influence relative to their publication counts. These patterns show that influential contributions in the field are not limited to authors with the highest publication output.
The m-index further highlights authors whose work has accumulated citations rapidly relative to the duration of their activity within the dataset. T.K.F. Chiu records the highest m-index (1.200), while several other authors, including Y. Wang, M.S. Ramírez-Montoya, C.D. Duong, X. Zhang, and B.L. Moorhouse, also show comparatively strong citation growth rates relative to the years since their first publication in the corpus.
Overall, the findings indicate that scholarly influence in AI and transversal skills research is unevenly distributed across authors and cannot be explained solely by publication volume. Instead, the field appears to be shaped by a combination of sustained productivity and individual contributions that attract substantial citation attention, reflecting its evolving and interdisciplinary nature.
Most Globally Cited Documents
Top 10 Most Globally Cited Documents (2020–2025)
Note. Total Citations = global citation count across all indexed databases; TC per Year = total citations divided by the number of years since publication; Normalized TC = citations normalized against the mean citation count for documents published in the same year within this dataset. Data sourced from Biblioshiny (bibliometrix R package).
Chan and Hu’s (2023) article, “Students’ Voices on Generative AI,” published in the International Journal of Educational Technology in Higher Education, is the most cited document, with 1,296 citations and a TC per Year of 324.00, which is nearly three times the annual rate of the second-ranked paper. Its Normalized TC value of 25.21 further indicates citation performance well above the average for publications from the same year. Chan also appears a second time in the top 10 with the paper “The AI Generation Gap” (Chan & Lee, 2023), which accumulated 398 citations. Together, these two publications account for 1,694 citations, reflecting strong scholarly visibility.
Several other documents also demonstrate substantial scholarly visibility. Southworth et al. (2023), “Developing a model for AI across the curriculum: Transforming the higher education landscape via innovation in AI literacy,” records 438 citations, while Thornhill-Miller et al. (2023), “Creativity, critical thinking, communication, and collaboration: Assessment, certification, and promotion of 21st century skills for the future of work and education,” reports 405 citations. Further, Elendu et al. (2024) medical education article, “The Impact of Simulation-Based Training in Medical Education: A Review,” accumulated 412 citations and the second-highest Normalized TC value (23.20), indicating rapid citation accumulation within a short period after publication. In addition, Broo et al. (2021), “Rethinking Engineering Education at the Age of Industry 5.0,” and Tapalova and Zhiyenbayeva (2022), “Artificial Intelligence in Education: AIEd for Personalized Learning Pathways,” illustrate the growing integration of AI-related transversal skills research across engineering, workforce development, and educational innovation.
The relatively high citation counts of several recent publications likely reflect the increased scholarly attention directed toward AI technologies following the widespread emergence of generative AI after 2022 (Gusteti et al., 2024; Irwanto, 2025). Research on AI and transversal skills intersects with multiple active research domains, including higher education, workforce development, digital transformation, and educational technology, which may facilitate rapid citation diffusion across disciplines. In emerging research areas, conceptual, review-based, and framework-oriented studies often accumulate citations quickly because they provide early theoretical interpretations, policy insights, and practical guidance for subsequent research. The citation patterns observed in Table 4 therefore align with the rapid development and interdisciplinary nature of the field.
Author Collaboration Network
The author collaboration network was generated using VOSviewer based on co-authorship analysis of the Scopus dataset (n = 1,487). A minimum threshold of three documents per author was applied, resulting in 43 authors meeting the criterion. Co-authorship networks are commonly used to map the social structure of a research field by revealing patterns of collaboration, connectivity, and institutional relationships among researchers (Zupic & Čater, 2015). However, as shown in Figure 3, the network demonstrates a relatively dispersed collaboration structure. The largest cluster (Red color) contains a relatively small group of collaborating authors, including Del Maestro, Rolando F., Fazlollahi, Ali M., Ledwos, Nicole, Yilmaz, Recai, and Alsayegh, Ahmad, while several productive authors appear as isolated nodes. This pattern is consistent with small and localized research teams, as well as intra-national rather than broad international collaboration. Given the relatively recent and still limited publication base, such fragmentation is not unexpected, since large and densely connected collaboration structures typically emerge over longer periods of sustained research activity (Zupic & Čater, 2015). The collaboration map should therefore be interpreted as reflecting an early-stage and institution-centered research network rather than a mature global collaboration structure. Author Collaboration Network in AI and Transversal Skills in Higher Education Research (2020–2025)
Given the inherent limitations of co-authorship analysis in emerging research fields, author co-citation analysis (Figure 4) provides a more informative perspective on the intellectual structure and developing research fronts of the field. Co-citation analysis maps relationships among authors who are frequently cited together and therefore reveals shared conceptual foundations even when direct collaboration is limited. Figure 4 thus complements Figure 3 by identifying influential authors and the intellectual foundations upon which current research is being developed. Author Co-Citation Network in AI and Transversal Skills in Higher Education Research (2020–2025)
Author Co-Citation Analysis
An author co-citation analysis was conducted using VOSviewer. A minimum threshold of five citations was applied, resulting in 142 authors meeting the criterion. As shown in Figure 4, the co-citation map identifies authors who are frequently cited together across the corpus and therefore reveals the intellectual structure and emerging research fronts underlying work on AI and transversal skills. Node size corresponds to co-citation frequency, edge strength reflects how often pairs of authors are cited together, and color clusters indicate groups of authors forming thematic or conceptual communities.
The network is strongly centered around Chiu TKF, who emerges as the most prominent node. This suggests that scholarship related to AI literacy and educational technology functions as an important intellectual reference point within the field. The pattern is also consistent with the productivity and citation impact identified in Table 3.
Surrounding Chiu are several distinct clusters. One cluster, including Kung and Chen, is associated with simulation-based learning and medical education. Another cluster, including Brown, Davis, Barrot, and Bearman, reflects digital pedagogy, AI-enabled learning, and competency development, suggesting that researchers across different applied domains are drawing on a shared set of influential works even when direct co-authorship is limited.
Interpreting the co-citation map alongside the author collaboration map (Figure 3) highlights an important distinction. While co-authorship reveals localized and institution-centered collaboration patterns, co-citation reflects broader intellectual convergence within the field. The co-citation clusters indicate that, despite fragmented co-authorship, scholars are referencing a common literature and building on related conceptual foundations. This makes co-citation analysis particularly useful for identifying research fronts and influential scholarly communities within an emerging interdisciplinary field.
Overall, the co-citation network demonstrates that, despite the fragmented collaboration patterns observed in Figure 3, the field is grounded in a relatively interconnected intellectual structure. This suggests that researchers are drawing upon a shared body of influential scholarship, indicating the emergence of common conceptual foundations even in the absence of extensive direct collaboration.
Further, Lotka’s law of author productivity was applied to examine the distribution of author contributions (Figure 5). Lotka’s law posits that the number of authors who have written n publications is proportional to 1/n2, meaning that a large proportion of authors publish only one paper, while progressively fewer authors publish multiple works (Lotka, 1926). This is a bibliometric principle commonly used to explain patterns of scientific productivity (Mittal et al., 2026). Author Productivity Distribution Based on Lotka’s Law
As Figure 5 illustrates, 93.4% of authors (5,229 of 5,597) published only one paper, compared with the 64.9% predicted by Lotka (1926). The theoretical value of 64.9% represents the proportion of authors expected to publish a single paper. The observed distribution (solid line) is much steeper than the theoretical distribution (dashed line), with an approximately 28.5-percentage-point difference at the one-paper level. This indicates that research on transversal skills in the age of AI in higher education is still an emerging field in which many researchers contribute only once. Such a pattern is typical of rapidly developing research areas that attract interest from scholars across multiple disciplines. As the field matures, this distribution may gradually shift toward the theoretical pattern predicted by Lotka’s law.
While author-level collaboration patterns reveal how individual researchers interact within the field, examining country-level production and citation impact provides a broader perspective on the geographic distribution and global influence of research on transversal skills in the age of AI.
Country Production and Citation Analysis
Figures 6 and 7 illustrate the relationship between regional publication volume and scientific impact. Accordingly, Figure 6 shows the cumulative publication growth trends of the five most productive countries from 2020 to 2025 in this dataset. All five countries demonstrate a clear increase in publications, with growth accelerating noticeably after 2023, when generative AI became a major global focus. The United States leads with 643 articles in 2025, followed by China (416), the United Kingdom (189), Spain (147), and Canada (143). Country-Wise Publication Trends Over Time (2020–2025) Top Countries by Citation Count (2020–2025)
China’s growth is particularly sharp, with its 2025 output nearly three times higher than in 2024, indicating a rapid expansion of research in this area. The United States and the United Kingdom show more steady but still strong increases. Although Spain and Canada have lower overall totals, both nearly doubled their publications between 2024 and 2025. Overall, the post-2023 rise across all five countries suggests that generative AI intensified research attention to transversal skills in higher education.
While country wise publication trends indicate research productivity, citation analysis provides additional insight into scholarly influence and research visibility. Figure 7 presents total citation counts by country and reveals important differences between publication volume and citation impact. Hong Kong leads in total citations (2,930), followed by China (2,448) and the United States (1,751). These findings suggest that countries with high research productivity also tend to accumulate substantial scholarly influence within the field.
However, total citation counts should be interpreted cautiously because larger publication outputs naturally tend to generate higher citation totals. Accordingly, citation impact relative to productivity, such as average citations per document, offers a more balanced perspective on scholarly influence. When this indicator is considered, Hong Kong (86.2), Nigeria (64.4), and Ireland (46.0) emerge as the leading countries in citation efficiency. This indicates that although China and the United States produce larger volumes of research, countries such as Hong Kong and Nigeria achieve comparatively stronger average impact per publication. Canada (26.0), the United Kingdom (25.6), and Germany (20.4) maintain a more balanced profile across both productivity and citation impact measures.
Overall, the findings demonstrate both the global expansion of research on transversal skills in the age of AI and notable differences in national research influence. Some countries dominate in publication volume, whereas others achieve stronger citation impact relative to their research output, highlighting varying patterns of scholarly contribution across the international research landscape.
Most Productive Affiliations
Figure 8 highlights the institutions contributing most actively to the field. The Chinese University of Hong Kong leads with 22 publications, followed by the Institute for the Future of Education with 18, and both the University of Florida and the University of Toronto’s Faculty of Medicine with 17 each. Hong Kong stands out in particular: three of its universities, the Chinese University of Hong Kong (22), the Education University of Hong Kong (16), and the University of Hong Kong (13), appear among the top seven, together accounting for 51 publications. This institutional clustering mirrors the strong citation influence of Hong Kong observed in the country-level analysis. Top Contributing Affiliations (2020–2025)
Medical education is also a major locus of activity. Several leading institutions in this area rank within the top 25, including the University of Toronto Faculty of Medicine (17), NYU Grossman School of Medicine (13), Harvard Medical School (10), and Emory University School of Medicine (9). Their presence underscores the central role of health professions education in shaping research on transversal skills in the AI era. The distribution of top contributing institutions across Asia, the Middle East, Latin America, North America, and Europe further reflect the global and interdisciplinary character of this research landscape.
Keyword Co-Occurrence and Temporal Evaluation
To address RQ3 and RQ4, a keyword co-occurrence network was generated using VOSviewer based on the Scopus dataset (n = 1,487). Keyword co-occurrence analysis reveals the conceptual structure of a research field by identifying how frequently key terms appear together across publications. Examining these patterns provides insight into the transversal skills most frequently discussed in relation to AI as well as the dominant and emerging research themes within the literature.
Prior to analysis, a thesaurus file was applied to merge keyword variants (e.g., “AI literacy” and “AI literacy”, “generative AI” and “generative artificial intelligence”), while non-substantive terms, including country names and database indexing terms, were removed to improve network interpretability. A minimum keyword occurrence threshold of five was applied. The network was constructed using VOSviewer’s default association strength normalization method, with clusters identified through the built-in modularity-based clustering algorithm, which groups keywords according to the strength and density of their co-occurrence relationships across the dataset (van Eck & Waltman, 2010). In the resulting network, node size represents keyword frequency, edge thickness indicates the strength of co-occurrence links between terms, and colors denote distinct thematic clusters within the field.
Network Visualization
As shown in Figure 9, the keyword co-occurrence network comprises 85 keywords connected by 749 links, with a Total Link Strength (TLS) of 2,066. The network was constructed using a minimum keyword occurrence threshold of 5, a resolution parameter of 0.70, and a minimum cluster size of 3. The modularity-based clustering algorithm identified five distinct thematic clusters with the following sizes: Cluster 1 (N = 22), Cluster 2 (N = 20), Cluster 3 (N = 20), Cluster 4 (N = 15), and Cluster 5 (N = 8). Structural interpretation is therefore based on node size, link strength, and cluster membership, consistent with standard practice in bibliometric keyword co-occurrence analysis (van Eck & Waltman, 2010). Keyword Co-Occurrence Network Map
The network reveals a centralized topology anchored by “artificial intelligence,” which serves as the primary conceptual hub connecting all five clusters. The network density and the high number of inter-cluster connections suggest a cohesive research landscape where themes are closely interrelated rather than isolated. Clusters 1, 2, and 3 represent the core of the field, accounting for 73% of the network (62 of 85 keywords), while Clusters 4 and 5 represent specialized and emerging frontiers. These structural indicators reflect a highly integrated and cohesive research core rather than a fragmented landscape of isolated topics.
Thematic Clusters and Temporal Evolution
As shown in Figure 9, the modularity algorithm identified five distinct clusters representing the major thematic structures within the field, varying in thematic maturity and density. The temporal evolution of these clusters is further illustrated through the overlay visualization in Figure 10, in which keyword nodes are colored according to their Average Publication Year (APY), revealing how the field has evolved from foundational themes toward more recent and emerging areas of research. • Cluster 1 (Red): AI-Driven Curriculum Transformation and Workforce-Ready Transversal Skills
Examples of key terms: artificial intelligence, higher education, education, curriculum, digital literacy, competencies, skills, computational thinking, blended learning, gamification, labor market, self-regulation, readiness, ethics, critical thinking.
This is the largest and most densely connected cluster, representing the highest level of thematic maturity in the network. It centers on themes related to curriculum transformation and workforce-relevant transversal skills in AI-driven educational contexts. The frequent pairing of “curriculum” with terms such as “competencies,” “skills,” and “labor market” points to a strong emphasis on aligning learning outcomes with the demands of future employment. The prominence of “computational thinking” and “critical thinking” highlights their status as core competencies for navigating AI-augmented environments. Terms like self-regulation, readiness, and ethics signal growing attention to self-directed and ethically grounded dispositions as essential elements of responsible AI engagement. The presence of terms such as blended learning and gamification suggests growing interest in innovative pedagogical approaches for supporting transversal skill development. In terms of temporal evolution, the earliest research in this cluster (APY, 2020–2022) reflects the foundational phase of the field, where these themes provided the conceptual pillars upon which more recent AI-specific inquiry has been built. • Cluster 2 (Green): Pedagogical Innovation for Transversal Skill Development
Examples of key terms: educational innovation, active learning, collaboration, creativity, experiential learning, project-based learning, soft skills, teaching, complex thinking, sustainable development goals, industry 4.0, education 4.0, technology-enhanced learning.
This cluster reflects work on innovative teaching approaches aimed at developing transversal skills for future employment in AI-shaped industries. The strong co-occurrence of “soft skills,” “collaboration,” “creativity,” and “complex thinking” shows that researchers are examining how these abilities can be strengthened through learner-centered methods such as “active learning,” “experiential learning,” and “project-based learning.” The appearance of “Industry 4.0” and “Education 4.0” signals a clear alignment between educational innovation, and the skill demands of the fourth industrial revolution. The inclusion of “sustainable development goals” suggests that transversal skills for future work are increasingly being framed within broader commitments to sustainability and global citizenship. The presence of technology-enhanced learning indicates that AI and digital tools are being used to support pedagogical innovation and to create environments where transversal skills can be meaningfully developed. Temporally, this cluster corresponds to the foundational phase (APY, 2020–2022), reflecting early work on competencies, pedagogy, and skills that preceded the emergence of generative AI as a dominant research focus, providing the conceptual scaffolding on which more recent AI-specific research has been built. • Cluster 3 (Blue/Cyan): AI Literacy and Digital Competence as Core Transversal Skills
Examples of key terms: generative artificial intelligence, AI literacy, digital competence, self-efficacy, self-determination theory, large language models, metacognition, learning outcomes, academic integrity, attitude, feedback.
Positioned as a structural bridge, this cluster occupies an intermediate position in the network. It centers on AI literacy and digital competence as key transversal skills in the age of AI. “AI literacy” and “digital competence” stand out as the most prominent skills, positioned where technological understanding meets critical engagement with AI systems. Their strong links to “self-efficacy” and “self-determination theory” show that researchers are drawing on established motivational frameworks to examine how learners’ confidence and intrinsic motivation shape the development of AI-related competencies. “Metacognition,” the ability to monitor and regulate one’s own learning, emerges as another key transversal skill, especially for meaningful interaction with generative AI tools and large language models. The co-occurrence of “academic integrity” with “AI literacy” highlights the ethical dimension of this work, underscoring that responsible and honest use of AI is now considered an essential competency for future professional practice. The presence of learning outcomes and feedback within the cluster points to ongoing efforts to measure, support, and refine the development of these transversal skills.
Temporally, this cluster reflects the transitional phase (APY, 2023–2024), where research shifts from broad digital-skills discourse to explicitly AI-focused inquiry, with self-determination theory gaining prominence as a framework for understanding motivation in AI-mediated learning environments. • Cluster 4 (Yellow): Educator Capacity Building for AI-Ready Transversal Skills
Examples of key terms: AI competency, teacher training, teacher education, sustainable development, stem education, student engagement, pedagogical innovation, artificial intelligence in education.
A comparatively smaller and more specialized cluster, it highlights research related to educator capacity building for supporting transversal skill development in AI-enabled learning environments. The pairing of “AI competency” with “teacher training” and “teacher education” points to a growing line of research examining how educators themselves must build these skills, particularly AI-related competencies to effectively prepare students for future work. The presence of “STEM education” suggests that this conversation is especially active in science and technology fields, where AI integration is advancing quickly. The inclusion of “student engagement” reflects interest in how AI-competent educators can create learning environments that support deeper involvement in transversal skill development. Terms such as “pedagogical innovation” and “sustainable development” indicate that building educator capacity is being framed not only as a means of improving immediate teaching practice but also as part of a broader, long-term vision for sustainable educational transformation in an AI-driven environment. Temporally, this cluster corresponds to the emerging frontier (APY, 2024.5–2025), reflecting growing attention to educator capacity building as the field moves toward a more applied, workforce-aligned direction. • Cluster 5 (Purple): Information Ecosystems, Ethics, and Professional Readiness
Examples of key terms: information literacy, ethics, innovation, digital skills, self-regulation, readiness, professional competencies.
This is the smallest and most peripheral cluster, representing a specialized thematic frontier where information literacy meets ethical reasoning and accountability. It focuses on the broader ecosystem of transversal skills needed for professional “readiness” in AI-shaped workplaces. Information literacy stands out as a key skill that links traditional information practices with emerging expectations around AI literacy. The pairing of “ethics” with “professional competencies” highlights the growing recognition that ethical reasoning is an essential transversal skill, especially as AI introduces new challenges related to bias, transparency, and accountability. Terms such as “self-regulation” and “readiness” point to the importance of autonomous, self-directed dispositions that enable workers to adapt continuously in fast-changing, AI-driven environments. The presence of “innovation” within this cluster suggests that the ability to innovate, particularly by using AI as a creative and professional tool, is increasingly being framed as a core transversal skill for the future workforce.
Temporally, this cluster also corresponds to the emerging frontier (APY, 2024.5–2025), with the appearance of labor-market connections and professional readiness signaling that the field is moving toward more applied, workforce-oriented concerns. In this emerging phase, ethics is increasingly framed not only as a policy issue but also as a core transversal competency for responsible engagement with AI in professional contexts, while readiness and self-regulation highlight growing attention to the affective and adaptive dispositions required for future work in AI-driven environments.
The connection between the five clusters and their temporal emergence confirms that the research landscape is both conceptually coherent and responsive to technological shifts. The high degree of connectivity between the foundational core (Cluster 1) and the emerging frontier (Cluster 5) indicates that the research themes are closely interrelated rather than isolated. Strong interconnections between clusters, particularly between AI literacy (Cluster 3) and curriculum transformation (Cluster 1), as well as between pedagogical innovation (Cluster 2) and educator capacity building (Cluster 4), indicate that these themes are frequently examined in relation to one another.
Taken together, the five clusters illustrate how research on transversal skills in the age of AI is organized around several interconnected themes (Figure 10): • Curriculum transformation aimed at embedding workforce-ready transversal skills; • Innovative pedagogies designed to cultivate soft skills and higher-order thinking; • AI literacy and digital competence as foundational skills, often examined through motivational theories; • Educator capacity building to support AI-ready transversal skill instruction; and • Professional readiness shaped by information literacy, ethics, and self-regulation. Overlay Visualization Showing Research Trends
Density Visualization: Research Hotspots
Density visualization highlights areas where research activity is most concentrated. In this map, colors represent keyword density, with yellow areas indicating high concentrations of frequently co-occurring terms and blue areas indicating lower levels of research activity. As shown in Figure 11, the brightest region at the center of the map is dominated by “artificial intelligence,” extending toward “generative artificial intelligence,” “higher education,” “education,” and “curriculum.” These keywords represent the core topics structuring research on transversal skills in the age of AI. Keyword Density Map
Several secondary hotspots appear around this central region. One prominent area extends from “digital competence” toward “generative artificial intelligence,” indicating strong scholarly attention to AI-related literacies and competencies. Another moderately dense region around “assessment,” “critical thinking,” and “soft skills” reflects sustained attention to how transversal skills can be evaluated in AI-mediated learning environments. A third area centered on “competencies,” “digital literacy,” and “education” corresponds to established research on competency frameworks in higher education.
At the same time, several emerging areas appear in the lower-density peripheral zones of the map. Keywords such as “metacognition,” “self-regulation,” and “readiness” appear in relatively sparse areas, suggesting that these theoretically important skills have only recently begun to attract scholarly attention. Similarly, “ethics” and “information literacy,” although widely recognized as essential components of responsible AI engagement, remain less densely connected within the current research landscape.
Notably, the relationship between transversal skills and labor-market demands also appears in a low-density peripheral area, suggesting that connections between higher education research and workforce requirements remain underexplored. In addition, “computational thinking,” which appeared prominently in earlier phases of the field, now occupies a lower-density region, indicating that attention may be shifting toward newer AI-related competencies such as AI literacy, metacognition, and self-regulation.
Overall, density visualization suggests that research on transversal skills in the age of AI is organized around a core set of technology-related themes, particularly artificial intelligence, AI literacy, and digital competence. At the same time, several skills, including metacognition, self-regulation, ethics, and labor-market readiness, remain relatively underexplored and represent promising directions for future research. Taken together, the keyword co-occurrence, overlay, and density analyses reveal a research landscape increasingly centered on digital competence and emerging human-centered competencies required for AI-mediated environments.
Author Keyword Frequency Analysis
To complement the keyword co-occurrence analysis, a term-frequency analysis was conducted to identify the most frequently occurring author keywords in the dataset (Figure 12). The term-frequency analysis examines the most frequently occurring author keywords to identify the intellectual landscape of research on transversal skills in the age of AI in higher education. The term-frequency analysis shows that “artificial intelligence” (478 occurrences) is the most frequent, followed by “higher education” (230). Together, they account for more than one-third of all keyword occurrences and define the field’s central framing. “ChatGPT” (132) appears next, surpassing all transversal-skill terms and illustrating the strong influence of this tool. The wider AI-technology cluster, including “generative AI” (72), “machine learning” (48), and “deep learning” (31), reinforces the field’s technological orientation. Most Frequent Author Keywords
Among transversal skills, “AI literacy” (72) is the most frequent term, ahead of “digital literacy” (32), “digital competence” (28), and “critical thinking” (25), indicating a shift toward AI-specific literacies. Psychological constructs such as “self-determination theory” (26) and “self-efficacy” (21) highlight the motivational foundations of skill development. At the same time, “academic integrity” (15) and “employability” (14) reflect growing attention to ethical engagement and workforce readiness.
Educational contexts are also strongly represented. “Medical education” (67) and “nursing education” (17) together account for 84 occurrences, confirming the health professions as major contributors. “Engineering education” (21), “teacher education” (17), and “curriculum development” (19) further demonstrate the field’s disciplinary breadth. Overall, these patterns align with the five thematic clusters identified in the co-occurrence analysis discussed in the previous section.
Together, these patterns reinforce the central role of AI technologies, particularly generative AI, in shaping contemporary discussions of transversal skills in higher education.
Discussion
This study mapped the evolving research landscape of transversal skills in higher education in the age of AI through a bibliometric analysis of 1,487 Scopus-indexed publications between 2020 and 2025. The findings indicate a rapidly expanding research domain characterized by a sharp rise in publication output, increasing interdisciplinary engagement, and growing attention to competencies related to artificial intelligence. Together, these patterns suggest that generative AI has intensified scholarly attention to the competencies required for learning and working in AI-driven environments.
The sharp and sustained increase in publication output following 2022 reflects the transformative impact of generative AI on higher education. The public availability of tools such as ChatGPT from late 2022 onward intensified scholarly discussions on digital skills, assessment, and workforce preparation, rapidly reshaping research and practice (Irwanto, 2025; Polat et al., 2024). The annual growth rate of 94.95% mirrors similar publication surges reported in bibliometric studies of AI in education (Henukh et al., 2025; Kavitha et al., 2024; Mittal et al., 2026), indicating that generative AI has become a major turning point across multiple research communities. However, this rapid expansion also presents a challenge: research is growing faster than institutions can meaningfully translate it into curriculum design, policy, and teaching practice. This creates what scholars describe as an “awareness-to-practice gap,” where research findings may not translate into meaningful change in educational systems (Almisad & Aleidan, 2025). The World Economic Forum (2023) and the OECD (2021) have similarly reported that educational institutions often lag behind both research findings and labor-market shifts when updating graduate competency frameworks. Closing this gap requires institutions not only to produce and engage with research, but also to develop the organizational capacity needed to translate evidence into program-level action through mechanisms such as curriculum review cycles informed by emerging evidence on AI-related competencies, cross-faculty working groups on AI integration, and closer alignment between institutional policy and evolving graduate skill requirements (OECD, 2021). At a broader level, this also points to an important role for national qualification frameworks and accreditation bodies in supporting greater consistency in AI-related transversal competencies across institutions and disciplines.
The findings also highlight the interdisciplinary nature of this research field. Publications appear across a diverse range of outlets, including educational technology journals, medical and health professions education journals, and multidisciplinary platforms. The prominence of medical education journals reflects the long-standing tradition of competency-based training in health professions education and the rapid integration of AI diagnostic tools into clinical practice, which has accelerated research on AI-related skill development in this field. This pattern suggests that competency-based approaches may provide a useful framework for disciplines seeking to define and assess AI-related transversal competencies within their programs. At the global level, the United States and China emerge as the most productive countries, while regions such as Hong Kong demonstrate high citation impact despite comparatively smaller publication volumes. These patterns indicate that research on transversal skills in the AI era is shaped by diverse disciplinary and geographic contributions, while also highlighting the need for broader representation in future research.
Despite this interdisciplinary scope, co-authorship patterns indicate that collaboration networks remain relatively fragmented, with many researchers working within small institutional teams rather than extensive international partnerships. Similar patterns have been observed in studies of artificial intelligence in education (Li & Rohayati, 2025; Mittal et al., 2026), suggesting that the field is still structurally developing. These fragmented collaboration networks highlight the need for stronger cross-disciplinary and cross-institutional partnerships. Research on emerging fields consistently shows that structured international collaboration accelerates knowledge consolidation, reduces duplication, and produces findings with greater generalizability (Donthu et al., 2021; Zupic & Čater, 2015). At the same time, the author co-citation analysis suggests that the field is grounded in a shared intellectual foundation, with scholars such as Chiu, Brown, Kung, and Barrot serving as common reference points across disciplinary boundaries. This indicates that intellectual coherence is emerging even where direct collaboration remains limited.
The keyword analyses further reveal the expanding prominence of AI-related competencies within the literature. In particular, AI literacy and digital competence appear frequently alongside higher-order cognitive capabilities such as critical thinking, collaboration, self-efficacy, and metacognition. This pattern reflects growing recognition that meaningful engagement with AI-enabled environments requires both technological understanding and broader cognitive capabilities. These findings align with broader discussions in higher education policy and workforce research emphasizing the importance of adaptable and interdisciplinary skills in response to rapid technological change (OECD, 2021; World Economic Forum, 2023). Importantly, this shift suggests that higher education scholarship is moving beyond treating AI as merely a technical development and is increasingly attentive to the human-centered cognitive, social, and ethical capabilities needed for effective participation in AI-mediated contexts.
At the same time, the results suggest that several areas remain relatively underdeveloped within the research landscape. Topics such as metacognition, self-regulation, and ethical engagement with AI appear as emerging themes but remain comparatively less dense within the current literature. Similarly, the connection between transversal skills research and workforce demands remains only partially explored, with terms such as “employability” and “labor market” appearing only at the periphery of the keyword network. These patterns highlight the need to examine how higher education can better support learners in developing reflective, responsible, and workforce-relevant competencies in AI-shaped contexts. The relative underrepresentation of these themes also points to an important future direction for the field, particularly in relation to the higher-order and ethical dimensions of transversal skills.
These findings also point to substantial practical implications for higher education institutions seeking to prepare students for AI-driven work environments. First, the growing importance of AI literacy, critical thinking, and metacognition within the literature, alongside the emergence of ethical reasoning and self-regulation as growing areas of concern, suggests that meaningful graduate preparation requires more than exposure to AI tools. It requires deliberate and sustained attention to the broader competencies that enable students to use those tools thoughtfully, critically, and responsibly. Institutions that embed these competencies across curricula and assessment practices, rather than confining them to isolated modules, are likely better positioned to produce graduates who can navigate the demands of AI-driven workplaces with both confidence and judgment. Second, the appearance of assessment-related terms in the keyword analysis also highlights the need to redesign assessment practices. Traditional assessment methods are increasingly vulnerable to the widespread use of AI, creating greater demand for authentic, process-oriented, and competency-based approaches that capture students’ reasoning, ethical judgment, and metacognitive engagement.
The prominence of educator-focused themes in the co-occurrence analysis indicates that faculty development and capacity building is not only a future research priority but also an immediate institutional imperative. Sustained investment in faculty AI literacy, pedagogical training, and communities of practice is essential if curriculum changes are to be implemented meaningfully. The underexplored connection between higher education research and evolving workforce expectations suggests a need for more intentional engagement with employers and industry partners to ensure that educational approaches remain responsive to broader shifts in professional practice. Beyond these implications, the analysis also reveals several emerging directions that warrant further attention.
Emerging Trends and Future Directions in Transversal Skills Research
The temporal and density analyses reveal several emerging developments in research on transversal skills in the age of AI. One emerging trend is the increasing emphasis on AI-related literacies as key competencies for higher education and future work. Recent studies highlight the importance of enabling students not only to use AI tools but also to understand how AI systems function, critically evaluate AI-generated outputs, and engage with these technologies responsibly (Long & Magerko, 2020; Ng et al., 2024). This shift reflects a broader movement from general digital skills toward competencies that support effective human–AI collaboration in learning and working environments.
At the same time, the research landscape indicates that several theoretically important areas remain underexplored. In particular, metacognition, self-regulation, and ethical engagement with AI appear as emerging but relatively low-density topics in the current literature. These competencies are increasingly recognized as essential for meaningful interaction with AI systems, as they enable learners to monitor their thinking processes, critically assess AI-generated information, and make informed decisions when using AI tools (Ng et al., 2024; Weber et al., 2025). Further empirical research is therefore needed to examine how these higher-order skills can be effectively developed and assessed within AI-mediated learning environments.
Another important gap concerns the relationship between transversal skills and labor market demands. Although discussions of workforce readiness are beginning to appear in the literature, this connection remains relatively underdeveloped. Global reports such as the World Economic Forum (2023) highlight that skills such as analytical thinking, creativity, adaptability, and collaboration will become increasingly important in future AI-driven workplaces. However, research in higher education has not yet fully explored how these competencies align with evolving employer expectations. Future studies could therefore benefit from stronger engagement with labor market perspectives through multi-stakeholder research involving educators, policymakers, and industry partners.
Educator capacity building also represents an important area for future research. As artificial intelligence becomes increasingly integrated into higher education, educators play a critical role in designing learning experiences that foster transversal skills. Future studies should explore how faculty training, AI literacy development, and pedagogical innovation can support the effective integration of AI while helping prepare graduates for AI-driven future work environments.
Further opportunities for research also exist in connecting established educational research traditions with emerging AI-related competencies. For example, critical thinking assessment has a long-standing research tradition in education (Ennis, 2018; Facione, 1990), yet it remains relatively disconnected from recent discussions of AI literacy and metacognition. Integrating these areas could support the development of more holistic approaches to transversal skill development and evaluation in AI-enabled educational contexts.
Finally, comparative and longitudinal research may provide valuable insights into how transversal skills for the AI era develop across different educational systems and cultural contexts. Longitudinal studies that track the transferability of AI-related skills from higher education into professional practice would be particularly valuable for informing curriculum design and institutional strategies aimed at preparing graduates for AI-enabled workplaces.
Limitations
Several limitations should be acknowledged. The study relies exclusively on Scopus, and relevant studies indexed only in other databases may have been excluded. The restriction to English-language publications may introduce linguistic bias. In addition, the analysis was restricted to journal articles, which may underrepresent research disseminated through proceedings, book chapters, and other document types. Further, the bibliometric approach analyses structural patterns rather than substantive content and should be complemented by systematic reviews for deeper insight into specific competency areas. Keyword-based analysis depends on the consistency of author-assigned keywords across disciplines. Finally, given the field’s exponential growth, the landscape may have evolved further since the search date of February 2026.
Conclusion
This bibliometric study analyzed 1,487 Scopus-indexed publications from 2020 to 2025 to examine the development of research on transversal skills in higher education in the age of artificial intelligence. The findings reveal a rapidly expanding research landscape, particularly following the emergence of generative AI, which has intensified scholarly attention to the competencies required for learning and working in AI-enabled environments.
The analysis identifies AI-related literacies and digital competence as central areas of research attention, while competencies such as critical thinking, collaboration, metacognition, and ethical reasoning are receiving growing visibility. These patterns suggest that research on transversal skills is evolving beyond general digital skill frameworks toward competencies that support effective human–AI collaboration. At the same time, areas such as workforce readiness, self-regulation, and ethical engagement with AI remain comparatively underexplored, highlighting important directions for future research.
Beyond mapping publication trends, this study contributes to understanding how higher education research is responding to broader technological and workforce transformations associated with AI. The findings highlight the importance of preparing students not only with technical knowledge, but also with the cognitive, ethical, and adaptive competencies needed for meaningful participation in AI-driven environments.
Overall, this study provides a comprehensive overview of the evolving intellectual structure and research priorities surrounding transversal skills in the age of AI. As generative AI continues to reshape higher education and professional practice, the development of transversal competencies is becoming an increasingly important priority for preparing graduates to engage responsibly and effectively in AI-driven workplaces.
Footnotes
Consent to Participate
There are no human participants in this article and informed consent is not required.
Funding
The author received no financial support for the research, authorship, and/or publication of this article.
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.