Generative Artificial Intelligence and Transversal Competencies in Higher Education: A Systematic Review

Abstract

Generative AI is an emerging tool in higher education; however, its connection with transversal competencies, as well as their sustainable adoption, remains underexplored. The study aims to analyze the scientific and conceptual development of generative artificial intelligence in higher education to identify the most relevant transversal competencies, strategic processes for its sustainable implementation, and global trends in academic production. A systematic literature review (PRISMA) was conducted on the Web of Science, Scopus, and PubMed, analyzing 35 studies for narrative synthesis and 897 publications for bibliometric analysis. The transversal competencies identified were: Academic Integrity, Critical Thinking, Innovation, Ethics, Creativity, Communication, Collaboration, AI Literacy, Responsibility, Digital Literacy, AI Ethics, Autonomous Learning, Self-Regulation, Flexibility, and Leadership. The conceptual framework connotes the interdisciplinary nature and five key processes were identified to achieve the sustainable integration of Generative AI in higher education oriented to the development of transversal competencies: (1) critical and ethical appropriation, (2) institutional management of technological infrastructure, (3) faculty development, (4) curricular transformation, and (5) pedagogical innovation. On bibliometric behavior, scientific articles predominate, with few systematic reviews. China leads in publication volume, and social sciences are the most prominent area. It is concluded that generative artificial intelligence is key to the development of transversal competencies if it is adopted from a critical, ethical, and pedagogically intentional approach. Its implications and future projections in the field of higher education are discussed.

1. Introduction

Generative Artificial Intelligence (Generative AI or GenAI) has documented evidence for more than a decade in major databases such as Web of Science, Scopus, and PubMed; however, its exponential growth and greater impact have been observed since 2023. In particular, this year has marked a turning point in the field of higher education, culminating in the most cited and influential studies in the field [1,2,3]; coincidentally, they all deal with ChatGPT, considered to be the most advanced chatbot in the world so far [4], while other research calls for a deeper look at the specific strengths and weaknesses of Generative AI including GPT-4 [5], analyzing multiple perspectives of its introduction in education [6].

Precisely one influential study bringing together 43 expert contributions highlights divided opinions on ChatGPT regulation and points to key areas for further research: knowledge, transparency, and ethics; digital transformation; and education, learning, and academic research [2].

In the current landscape of generative artificial intelligence, ChatGPT has acquired a notable presence in the educational field due to its accessibility and widespread use. Although the review does not focus exclusively on this tool, its prominence in the literature justifies its frequent appearance as a representative example. Together with ChatGPT, other relevant Generative AI is considered to provide a broad view of their impact on the development of transversal competencies in higher education.

The integration of Generative AI in higher education is transforming didactic processes, providing personalized recommendations, fostering collaboration and communication, and improving learning outcomes [7]. In addition, these systems provide cognitive and emotional support, improve accessibility, and create inclusive learning environments [8]. However, despite its benefits, the implementation of Generative Artificial Intelligence poses significant challenges. For example, studies report that generative AI can facilitate plagiarism and authorship impersonation, raising serious concerns about the authenticity of student work [9,10]. Another study suggests that excessive use of generative AI may diminish students’ ability to develop critical thinking and creativity skills, as they may become overly reliant on automated solutions [11]. At the same time, it is revealed that the lack of transparency in the sources used by these tools can also affect the reliability of the information [9].

A qualitative study in Mexico focused on ChatGPT as a new tool to strengthen transversal competencies in higher education students reported that “students not only learned to use the new AI tool and deepened their understanding of prospective methods, but also strengthened three soft or transversal competencies: communication, critical thinking, and logical and methodical reasoning” [12]. This evidence highlights the potential of Generative Artificial Intelligence to foster the development of transversal competencies essential for human development, and these possibilities, although promising, have been little explored in the educational field and require further research attention, especially when a previous study highlights critical capacity, creative capacity, communicative capacity, and collaborative capacity as necessary skills to adapt to a digitized educational environment [13]. All the above calls for reflection on the link between Generative Artificial Intelligence and transversal competencies in higher education. In doing so, it becomes essential to clarify certain terms that are often used interchangeably, such as transversal competencies, soft skills, and digital skills, yet refer to distinct constructs.

Soft skills are a set of abilities related to personality traits, motivations, attitudes, and preferences that are not adequately captured by traditional academic performance or IQ tests. These skills are highly valued in the labor market and life in general and are causally associated with personal and professional success in life [14].

In the context of graduate employability, soft skills are becoming increasingly important according to the perceptions of both students and employers. Differences in the valuation and prioritization of these skills are evident, underscoring the need for collaboration between educational institutions and companies to promote their development and continuous adaptation to the demands of the labor market. This study identifies several soft skills such as professional ethics, work commitment, stress tolerance, creativity and innovation, learning skills, work-life balance, self-awareness, communication skills, conflict management and negotiation, cultural intelligence, leadership, teamwork, networking, adaptability to change, decision-making, among others [15].

Following the analysis, digital skills refer to specific and technical abilities related to the use of technological tools, which require specialized training and ongoing updating. A study conducted by van Laar and collaborators [16] analyzes the relationship between 21st-century skills and digital skills through a systematic review of the literature. The authors conclude that 21st-century skills are a broader concept than digital skills, as they include a wider range of abilities that do not depend exclusively on information and communication technologies (ICT). The study identifies seven core digital skills: technical skills, information management, communication, collaboration, creativity, critical thinking, and problem-solving. In addition, it highlights five contextual skills that complement the digital framework: ethical awareness, cultural awareness, flexibility, self-direction, and lifelong learning. This comprehensive approach emphasizes that while digital skills are fundamental to competitiveness and innovation in the knowledge economy, they must be articulated with social, cognitive, and ethical skills to train adaptive and effective professionals in changing contexts.

The study by Reisoğlu and Çebi [17] conceptualizes digital competencies through the DigComp and DigCompEdu frameworks, highlighting their development via practice-based, collaborative training across key areas such as information and data literacy, digital content creation, safety, and problem-solving. This perspective is highly relevant to the present study on generative artificial intelligence (GenAI), as it underscores that AI literacy should not be approached solely from a technical standpoint but rather as a critical extension of digital competence—integrating ethical, pedagogical, and reflective dimensions essential for navigating complex and dynamic educational environments.

However, according to the study “Transversal competencies: Their importance and learning processes by higher education students” [18], transversal competencies in higher education are key skills that enable students to cope with uncertainty and rapid changes in the labor market and today’s society. These competencies not only integrate knowledge but also attitudes, expectations, and predispositions that are fundamental to professional and personal success. The study highlights that although higher education institutions recognize the importance of developing transversal competencies, there is a gap between students’ expectations regarding the competencies they wish to acquire and the actual training offered by education systems. In essence, transversal competencies are understood as skills, attitudes, and dispositions that transcend specific areas of knowledge, promoting comprehensive training that favors adaptability, continuous learning, and the ability to respond to changing and multidimensional contexts.

Competencies are an integrated set of knowledge, skills, values, and judgments that enable individuals to act effectively across diverse contexts. Within this framework, transversal competencies encompass not only soft skills—such as communication, empathy, and teamwork—but also general knowledge, attitudes, values, and cognitive abilities, including critical thinking and the capacity for learning how to learn. While soft skills primarily focus on personal and social attributes related to socio-emotional interaction, transversal competencies integrate cognitive, attitudinal, and evaluative dimensions that facilitate adaptation, continuous learning, and the resolution of complex problems in various professional and academic settings. In summary, although soft skills constitute a fundamental component of transversal competencies, the latter represents a broader concept encompassing multiple dimensions of human performance.

Transversal competencies encompass and integrate both dimensions—soft skills and digital skills—as well as knowledge, attitudes, and cognitive abilities, forming a broad set that allows students to adapt, learn continuously, and face complex contexts. The study on Generative Artificial Intelligence highlights the importance of combining technological innovation with the ethical and critical development of these skills to transform higher education in a sustainable way.

For this study, 15 systematic reviews were identified among studies extracted from the Web of Science, Scopus, and PubMed, addressing topics such as the general implications of generative AI in higher education, the use of ChatGPT in writing, feedback and assessment, applications in medicine, trends and challenges, and effects on skills such as cognitive thinking and programming (see Appendix A). However, none of this research delves into the conceptual framework and the available scientific productivity, let alone the transversal competencies associated with these studies.

Thus, this study is particularly relevant at a time when AI is revolutionizing different sectors; it is imperative to discuss the importance of achieving transversal competencies in higher education [18], understanding that it is this “transversality” where the main objective of university teaching is based [13] which requires us today more than ever to consider the role of Generative AI. Indeed, one of the missions of higher education is to foster transversal competencies such as critical thinking, creativity, problem-solving, and effective communication, which are essential for development, particularly in a rapidly changing environment where new technologies play a decisive role [19].

Some of the most common transversal competencies are analysis and synthesis, organization and planning, communication, collaboration, interpersonal skills, use of ICTs, decision-making, leadership, and problem-solving [13], but in a digitalized world, an innovative system is needed to promote these transversal competencies in universities [19], This requires a new pedagogical positioning that allows a more precise response to the development of transversal competencies such as responsibility, proactivity, autonomy, adaptability, resilience and transferability [18].

Another interesting study systematizes the transversal competencies in digital competence (information literacy, data literacy, communication, cooperation, digital content creation, security, problem-solving), innovation competence (creativity, critical thinking, initiative, teamwork, networking), and research competence (attitude, ethics, research management, collaboration, and communication) [19].

Thus, in a globalized and technological world, transversal competencies have become a common theme in the field of research [20], but there is still a gap in their connection with the Generative Artificial Intelligence that has become part of our everyday life. Given this, the following objectives are proposed:

General Objective:

Analyze the scientific and conceptual development of generative artificial intelligence in higher education to identify the most relevant transversal competencies, strategic processes for its sustainable implementation, and global trends in academic production.

Specific Objectives:

Identify the most relevant transversal competencies associated with the study and use of generative artificial intelligence in higher education environments.

Analyze the conceptual structure of research on generative artificial intelligence in higher education and identify strategic processes that promote its sustainable development in this field.

To characterize the scientific production of generative artificial intelligence in higher education in terms of the type of document, geographical distribution, and thematic areas.

2. Materials and Methods

The PRISMA 2020 Statement has been used for this review [21]. strictly complying with the PRISMA 2020 Checklist (see Supplementary Materials). This allowed us to develop a systematic literature review as a strategy to identify the most relevant studies on generative artificial intelligence in higher education, considering that these reviews are used to identify, interpret and evaluate the data available in scientific production in a given period [22]. Operationally, the study followed a systematic review process consisting of five phases [23], specifying the PRISMA 2020 methodological criteria—eligibility criteria, sources of information, search strategy, study selection process, data extraction process, list of extracted data, assessment of risk of bias of individual studies, synthesis methods, assessment of publication bias, and assessment of certainty of evidence [21]. It is worth mentioning that the effect measurement criterion does not apply in this study.

Phase 1: Research questions (RQ). They are organized around three areas: (a) Conceptual Framework, to analyze the relationships between keywords identified in the literature (RQ1); (b) Bibliometric Performance, to identify Types of Documents, Geographic Location, Thematic Areas, and Words’ Frequency over Time (RQ2–RQ5); and (c) Transversal competencies (RQ6), to categorize the transversal competencies associated with generative artificial intelligence studies in higher education (see

Table 1. Research questions.

Systematization Criteria	Research Questions	Procedure
Transversal Competencies	RQ1: What are the Transversal Competencies associated with generative artificial intelligence studies in higher education?	Identification of competencies through theoretical systematization
Conceptual Framework	RQ2: What is the conceptual structure of research on generative artificial intelligence in higher education and strategic processes that promote its sustainable development in this field?	Analysis of co-occurrences of key terms and concepts in VOSviewer and Coding in Atlas.ti
Bibliometric Performance	RQ3: What is the distribution of scientific production by type of document?	Classification by document type
	RQ4: How does the geographical distribution of publications behave?	Identification of the country of affiliation of the first author and visualization through maps (Bibliometrix)
	RQ5: What thematic areas stand out in the scientific production on the use of generative AI in higher education?	Grouping by academic disciplines
	RQ6: What are the most frequent terms associated with generative AI in higher education, and how has their frequency of occurrence in the scientific literature varied between 2023 and 2025?	Words’ Frequency over Time (Bibliometrix)

).

Phase 2: Eligibility criteria and sources of information. We included articles published in Web of Science (WoS), Scopus, and PubMed between 2023 and 2025 that contain in their title the descriptors used in the search equation presented below, in English or Spanish. It is revealing that, despite not having executed temporal restrictions in the search, all the studies found date from 2023 onwards.

The eligibility criteria were mainly based on the elements of the study’s PICO Question: What are the transversal competencies associated with studies on the use of generative artificial intelligence tools in higher education, considering professors, students, and institutions?

-

Population: Teachers, students, or institutions of higher education.

-

Intervention: Use of generative artificial intelligence tools.

-

Results: Most relevant transversal competencies associated with the study and use of Generative AI in higher education.

The inclusion criteria for this study focus on research conducted with higher education teachers and students, covering universities and undergraduate, graduate, or postgraduate programs, regardless of whether the institutions operate under hybrid, face-to-face, or virtual models. Studies investigating the use of generative artificial intelligence tools, such as large language models (LLMs), text or image generators, and generative adversarial networks (GANs), assessing transversal competencies are considered if verifiable qualitative or quantitative results are provided. Empirical articles and systematic reviews published in indexed academic journals, in English or Spanish, from 2023 onwards are included, given the novelty of the line of research. As exclusion criteria, especially for the final synthesis, studies focused on educational levels other than higher education, those that do not use generative AI tools, publications in languages other than English or Spanish, documents other than articles or reviews such as Editorial, Note, Letter, Erratum, Conference paper, Book chapter, Book, and studies published before 2023 are discarded.

Phase 3: Search strategies. In this phase, we present the search equation, keywords, and combinations of terms using Boolean operators (AND, OR, NOT) and applied filters. For the search, location, and selection of studies in Web of Science (Wos), Scopus, and PubMed, a replicable search equation was designed, consisting of two thematic clusters; each cluster was linked internally with the Boolean operator “OR” and both clusters were linked together by the Boolean operator “AND”, limiting the search field to the title of the article (see

Table 2. Search equation.

Thematic Clusters	Search Equation
Generative Artificial Intelligence	“Generative Artificial Intelligence” OR “Generative AI” OR “Deep generative models” OR “Machine learning generative” OR “AI generation” OR “Generative Adversarial Networks” OR “Variational Autoencoders” OR “Natural Language Generation” OR “Large Language Models” OR “GenAI” OR “LLMs” OR “LLM” OR “GPT-4” OR “ChatGPT”
Boolean Operator	AND
Higher Education	“Higher education” OR “university” OR “undergrad” OR “bachelor’s degree” OR “postgraduate” OR “tertiary education”

Note: In PubMed, each descriptor was preceded by the title field as a setting of the platform’s search algorithm, for example: Generative Artificial Intelligence [Title].

).

Phase 4: Study selection process. This phase specifies the entire selection process, data extraction, and list of included data, considering the elimination of duplicates, review by title and abstract to discard irrelevant studies, review of the full text of the pre-selected studies to confirm their eligibility, recording the number of excluded studies and the reasons for their exclusion. Details are provided below (see

Table 3. Summary of articles by database.

Database	Total Documents	2025	2024	2023	ChatGPT Manuscripts	ChatGPT Percentage
Web of Science	283	3	223	57	165	58.3%
Scopus	574	12	449	113	318	55.4%
PubMed	40		32	8	30	75%
Total	897

Note: The date of the cut-off was 4 January 2025.

).

Phase 5: Data coding and synthesis. For data collection, the Zotero bibliographic manager was used, exporting all references in RIS and BibTeX format. The process involved various software and platforms, including R version 4.4.3, Bibliometrix version 4.3.2 [24], VOSviewer version 1.6.18, ATLAS.ti version 8, as well as the updated web versions of Covidence, and Rayyan.

The risk of bias assessment aimed to determine whether the design, execution, and analysis of the studies included in the final synthesis introduced biases that could compromise the validity of the findings. To this end, the CASP (Critical Appraisal Skills Programme) tool was used, addressing aspects such as the clarity of the objectives, methodological adequacy, clarity of sampling, validity of data collection and analysis, the reflexivity of the author, and the soundness of the conclusions.

Each criterion was evaluated using “yes”, “no”, or “partial” responses, finally issuing an overall judgment on the level of risk of bias (low, moderate, or high) of each study [25]. The CASP tool was used with minor modifications like GRADE [26].

The evaluation of the certainty of the evidence was carried out using the GRADE (Grading of Recommendations, Assessment, Development, and Evaluation) approach, which made it possible to grade the quality of the evidence according to key criteria: methodological limitations, consistency of the results, precision of the estimates and direct applicability of the evidence to the context of interest [27].

For the assessment of publication bias, two key aspects were considered: geographic and cultural diversity and completeness of sources. It was verified that the selected studies came from a variety of regions and contexts, ensuring a broad and balanced representation of the phenomenon under study. In addition, an exhaustive search strategy was employed, combining various internationally recognized databases, such as Web of Science, Scopus, and PubMed, to minimize possible biases and ensure the inclusion of a greater number of relevant studies. As a synthesis method, narrative synthesis was used to integrate the results of heterogeneous studies on generative AI in higher education. This approach allowed us to analyze and interpret the transversal competencies associated with the different studies in a structured manner.

3. Results

The following flow chart illustrates the complete process of identification, screening, and inclusion using Item 16a of the PRISMA guidelines as part of the results section (see Figure 1).

Of the 35 studies included in the final synthesis, 27 were rated as having a low risk of bias, indicating strong methodological design, rigorous execution, and appropriate data analysis. The remaining 8 studies were assessed as having a moderate risk of bias, which notably coincided with those that received the fewest citations. Item 16b of the PRISMA guidelines was not applicable, as all 35 studies that met the inclusion criteria were ultimately included in the final synthesis; therefore, no eligible studies were excluded at this stage.

Table 4. Selected studies.

No.	Author	Reference	Year	DOI	Risk of Bias	Citations in Scopus *
1.	Acosta-Enriquez et al.	[28]	2024	10.1007/s40979-024-00157-4	Low	24
2.	Al-Abdullatif	[29]	2024	10.3390/educsci14111209	Low	5
3.	Al-Saiari et al.	[30]	2024	10.48161/qaj.v4n3a760	Moderate	2
4.	Bannister, P.	[31]	2024	10.22055/rals.2024.45862.3214	Low	4
5.	Barrientos et al.	[32]	2024	10.59429/esp.v9i9.2927	Low	2
6.	Cano & Nunez	[33]	2024	10.3389/feduc.2024.1483853	Moderate	0
7.	Chen et al.	[34]	2024	10.1108/ILS-10-2023-0160	Low	11
8.	Consuegra-Fernández	[35]	2024	10.6018/rie.565391	Moderate	2
9.	Dabis & Csáki	[36]	2024	10.1057/s41599-024-03526-z	Low	19
10.	de Vicente-Yagüe-Jara et al.	[37]	2023	10.3916/C77-2023-04	Low	40
11.	Dhamija & Dhamija	[38]	2025	10.32674/ptf9yd75	Moderate	4
12.	Dillon	[39]	2024	10.15858/engtea.79.3.202409.123	Moderate	1
13.	Elbaz et al.	[40]	2024	10.1016/j.caeai.2024.100324	Moderate	3
14.	Fleischmann	[41]	2024	10.1080/14703297.2024.2427039	Low	2
15.	Fuchs & Aguilos	[42]	2023	10.18178/ijiet.2023.13.9.1939	Low	30
16.	Gasaymeh et al.	[43]	2024	10.3390/educsci14101062	Low	6
17.	Gong	[44]	2023	10.13366/j.dik.2023.05.097	Low	8
18.	Isiaku et al.	[45]	2024	10.53761/7egat807	Low	5
19.	Karkoulian et al.	[46]	2024	10.1007/s10805-024-09543-6	Low	3
20.	Kirwan	[47]	2024	10.1080/03323315.2023.2284901	Low	10
21.	Moongela et al.	[48]	2025	10.1007/978-3-031-78255-8_21	Moderate	0
22.	Nazari & Saadi	[49]	2024	10.1007/s44217-024-00122-w	Low	8
23.	Nikoçeviq-Kurti & Bërdynaj-Syla	[50]	2024	10.22521/edupij.2024.133.2	Moderate	3
24.	O’Dea et al.	[51]	2024	10.1177/14782103241287401	Low	6
25.	Raman et al.	[52]	2024	10.1155/2024/3085910	Low	26
26.	Rehman et al.	[53]	2024	10.1016/j.techsoc.2024.102655	Low	9
27.	Ruiz-Rojas et al.	[54]	2024	10.3390/su16135367	Low	29
28.	Sandu et al.	[55]	2024	10.1007/s44217-024-00126-6	Low	15
29.	Sevnarayan	[56]	2024	10.33902/JPR.202426525	Low	7
30.	Shahzad et al.	[57]	2024	10.1007/s10639-024-12949-9	Low	27
31.	Tseng & Lin	[58]	2024	10.34190/ejel.21.5.3329	Low	19
32.	Van Horn	[59]	2024	10.55593/ej.28109a8	Low	10
33.	Wang et al.	[60]	2024	10.1080/10447318.2024.2383033	Low	56
34.	Xu et al.	[61]	2024	10.1371/journal.pone.0295646	Low	19
35.	Yusuf et al.	[62]	2024	10.1186/s41239-024-00453-6	Low	99

* Note: The last cut-off date for the citation review was 2 June 2025.

provides a comprehensive summary of the studies included in the final synthesis using Items 17, 18, and 19 of the PRISMA guidelines. Specifically, the table cites each included study and outlines its main characteristics (Item 17), presents the assessed risk of bias categorized as low, moderate, or high (Item 18), and highlights the key findings relevant to the objectives of this review (Item 19). This synthesis facilitates a clear comparison across studies and supports the interpretation of the overall evidence base.

RQ1. What are the transversal competencies associated with generative artificial intelligence studies in higher education?

The transversal competencies identified in the selected studies are presented below (see

Table 5. Transversal competencies addressed in the reviewed studies on Generative AI in higher education.

Transversal Competencies	Key Findings/Evidence	Reference(s)
Academic Integrity	Emphasis on responsible use of AI and policies to prevent misconduct	[28,42]
Critical Thinking	GenAI improves critical, reflective, creative, and computational thinking.	[44,50,54]
Innovation	GenAI adoption is linked to innovation attributes and positive perceptions.	[29,48]
Ethics	Ethics as a mediating factor in AI’s educational use and student performance	[24,36,41,53,56]
Creativity	GenAI supports idea generation, creative writing, and verbal creativity	[33,37,41,53]
Flexibility	Use of GenAI linked to flexible learning strategies and adaptation	[32,33,51]
Communication	ChatGPT promotes effective discourse, engagement, and connectedness	[31,45,52]
Collaboration	Peer-to-peer, student-teacher, and international collaboration enhanced by GenAI	[26,34,39,50,54,55]
Metacognition & Self-regulation	GenAI supports self-efficacy, motivation, perseverance, and self-regulated learning.	[35,38,57]
AI Literacy	AI literacy is essential for effective, ethical, and responsible GenAI use.	[25,30,40,47,56]
Responsibility	Responsible AI use promotes learning outcomes and academic ethics	[36,43,56,58]
Educational Leadership	GenAI enhances leadership in academic settings by optimizing time and resources.	[46,56]

).

The study on generative artificial intelligence in higher education underscores the significance of transversal competencies—skills that are essential across diverse academic and professional settings. These encompass critical thinking, communication, teamwork, leadership, ethics, and creativity, along with more specialized areas such as digital literacy and AI ethics. Particularly noteworthy are competencies like academic integrity, innovation, and communication, which highlight the importance of a responsible, ethical, and reflective approach to the integration of AI in educational contexts (see Figure 2).

This result legitimizes that curricular design in higher education must integrate transversal competencies that enable students to develop ethically, critically, and creatively in an environment increasingly influenced by generative artificial intelligence. Academic integrity and ethics must be fundamental pillars to ensure the responsible use of these technologies, while critical thinking, innovation, and creativity drive the ability to adapt and solve problems in changing scenarios. Communication, collaboration, and leadership foster work in multidisciplinary teams, which is essential for AI development and governance. In addition, AI and digital literacy are key to the understanding and effective application of these tools in different fields of knowledge. Responsibility, autonomous learning, self-regulation, and flexibility strengthen student autonomy, allowing them to evolve along with technological advances. Thus, a curriculum design that contemplates these competencies prepares students for their professional future, promoting a comprehensive education aligned with the challenges of the digital era.

Per PRISMA Items 20a–20d, the results of this review are presented through a narrative synthesis. First, for each synthesis (Item 20a), we briefly summarize the characteristics of the contributing studies along with their respective risk of bias assessments. As no meta-analysis was conducted, statistical syntheses and summary effect estimates (Item 20b), as well as analyses of heterogeneity (Item 20c), are not applicable. Similarly, sensitivity analyses to assess the robustness of the synthesized results (Item 20d) were not performed. Instead, the narrative synthesis focuses on identifying patterns, consistencies, and variations across the studies to provide a comprehensive interpretation of the findings.

This part provides a narrative synthesis from the 35 studies that were included in the final selection, legitimizing the transversal competencies associated with Generative Artificial Intelligence studies in higher education. First, academic integrity is established precisely; a study with 24 science and mathematics students highlights the importance of balancing technological advances with the maintenance of academic integrity, emphasizing the need for responsible use of AI and the development of policies to mitigate potential negative impacts [32], another qualitative study, with higher education faculty and students at universities across Lebanon, found that while ChatGPT and AI tools are recognized for their potential to enhance productivity, promote interactive learning experiences, and provide personalized support, they also raise significant concerns regarding academic integrity [46].

This is aligned with the development of critical thinking as a competence that integrates logical and ethical reasoning and enables informed decision-making for problem-solving; in this regard, a study found that 64% of university students in their sample believe that Generative Artificial Intelligence tools significantly improve their critical thinking skills, although the authors emphasize the need for further research to better understand their long-term impact and maximize their potential benefits [54]. In the same line, another study evidenced the impact of Generative Artificial Intelligence on cognitive thinking skills identified as critical thinking, creative thinking, reflective thinking, computational thinking, and problem-solving [48].

On the topic of innovation, a study explores the adoption and social implications of an emerging technology such as Chat Generative Pre-Trained Transformer (ChatGPT) among higher education students, and the results suggest that five innovation attributes significantly impact adoption rates and perceptions of ChatGPT, indicating its potential for transformative social change within the education sector [52]. At the same time, another research study conducted a cross-sectional predictive analysis using data from 92 undergraduate business students at a Peruvian higher education institution and found a strong positive relationship between perceived enjoyment and intention to use Generative AI in innovation courses [33].

Regarding ethics and creativity, a study with 362 students from Chinese universities the findings revealed that generative AI technologies, such as LLM models exemplified by ChatGPT and chatbots, significantly influence students’ learning performance through self-efficacy, fairness and ethics, and creativity [57]. While another study with 201 university students reported that no sex or age differences were found in the relationship between ChatGPT use and perceived ethics, the study also found no differences in the relationship between ChatGPT use and perceived ethics [28]. Another study with 75 higher education academics demonstrates that ChatGPT can increase educational productivity, experience, creativity, and idea generation. However, it underscores the importance of careful consideration of ethical issues related to academic honesty and possible over-reliance on ChatGPT [45].

Related to flexibility and creativity, a study with 20 AI systems and 193 university students highlights the assistance provided by AI in writing tasks and verbal creativity and flexibility [37]. Data is also collected from 362 students in Chinese universities, reavealing that generative AI, such as LLM models exemplified by ChatGPT and chatbots, significantly influences students’ learning performance through self-efficacy, fairness and ethics, and creativity [57]. At the same time, empirical evidence was compiled from 30 leading universities ranked among the top 500 in the Shanghai Ranking list from May to July 2023, highlighting the role of flexibility in determining how they use generative AI, especially large language models, in their courses [36] while showing evidence of Generative Artificial Intelligence as a basic component for the creative process [41].

An interesting study highlights the importance of effective AI communication discourse in educational settings [49], while another study claims that ChatGPT can help learners cultivate non-cognitive skills, including motivation, perseverance, self-regulation, and self-efficacy, as well as metacognitive skills such as self-determination, self-efficacy, and self-regulation, by providing feedback, fostering creativity and stimulating critical thinking activities [61].

Concerning communication and collaboration, findings revealed that students found ChatGPT more engaging and interactive, feeling more connected to their peers and teachers [56]. While with 305, the results indicate that the interaction with ChatGPT increases freedom and productivity [53].

In an interesting study with 51 professors at the University of Barcelona specialized in communication and philology, subjects were presented with a sample of texts extracted from an authentic academic paper that included versions written by the students themselves, together with results generated ad hoc by ChatGPT, and there was a tendency for the texts generated by ChatGPT to be rated more favorably than those written by the students themselves [35].

This study with 121 university students showed a significant impact of generative artificial intelligence tools on critical thinking and collaboration [54]. The same is true in a qualitative study with 15 students where collaboration with GPT-3.5 was shown to foster critical thinking and enable students to develop a distinctive writing voice [58], demonstrating the creative potential of these tools within the educational environment to promote collaboration as demonstrated in a study with 95 undergraduate and graduate students [43], finding reinforced by research with 27 teachers and 409 students, with favorable results of Generative Artificial Intelligence in peer-to-peer collaboration [38], international collaboration as this study with 31 teachers demonstrates [30], The potential of ChatGPT in 120 students on their collaborative learning, metacognitive awareness and autonomous learning was evidenced [59]. Meanwhile, in a study with 10 teachers, the findings indicate that teachers see ChatGPT as a valuable tool for streamlining time-consuming teaching tasks, such as generating content ideas, creating engaging activities, and providing resources [50].

A key aspect revealed is AI literacy [44,60]. Studies provide empirical evidence on how better AI-literate college students use Generative AI more effectively [34], although the management of university policies regarding the ethical and appropriate use of generative AI remains a pending task [34], which in some studies is illustrated as practitioner discomfort with GenAI-assisted academic misconduct [31]. However, a study of 237 university teachers reveals that AI literacy is a key factor in teacher acceptance of GenAI [29]. This aligns with the results of other research with 234 students that showed that prior learning about AI has a large impact on Generative Artificial Intelligence literacy [51].

Another transversal competence revealed was responsibility, precisely a study with 1217 teachers from 76 countries found that the responsible use of generative AI tools improves learning processes [62]; however, another study discussing the introduction of ChatGPT to a group of undergraduate students, concludes that while the technology can offer assistance in completing the academic assessment, it does not replace higher thinking skills [47].

College students’ intentions to adopt ChatGPT increase their academic performance. Personal morality and ethics cause them to feel a sense of responsibility for their actions that violate academic integrity [40]; thus, students’ cognitive literacy in AI ethics and AI awareness literacy are crucial aspects. [60], This requires (1) support for autonomous learning, (2) digital and artificial tutoring, and (3) academic integrity training [42], harnessing the power of Generative Artificial Intelligence to promote self-regulated learning [39], flexibility [55], and educational leadership [60]. Thus, ethics in IA becomes one of the transversal competencies supported in several of the proposals [28,40,57,60].

Item 21 of the PRISMA guidelines, the risk of bias due to missing results, potentially stemming from publication bias, was assessed across the included studies. No significant concerns were identified, as the body of evidence appeared to be consistently reported and transparently documented.

In line with Item 22 of the PRISMA guidelines (certainty of evidence), the overall certainty in the body of evidence was rated as moderate to high based on criteria such as study design, consistency, directness, precision, and risk of bias. Of the 35 studies included, 27 showed a low risk of bias, indicating strong methodological quality, while the 8 with moderate risk had lower citation impact, suggesting a limited effect on the conclusions. No eligible studies were excluded, supporting the comprehensiveness and reliability of the synthesis.

RQ2: What is the conceptual structure of research on generative artificial intelligence in higher education and strategic processes that promote sustainable development in this field?

VOSviewer version 1.6.18 software was used to create the conceptual network, which allowed not only the identification of the most relevant concepts but also the association between them (see Figure 3).

The yellow cluster mainly addresses the ethical and educational competencies needed for the proper use of generative artificial intelligence, with emphasis on digital literacy and academic integrity; the purple cluster represents connections related to ethics and learning processes linked to the adoption and use of technologies such as ChatGPT, the green cluster focuses on the use of educational technologies and educational computing in higher education and how these influence behavioral intentions and student experience, the blue cluster with main concepts such as: “Large language model”, ‘language processing’, ‘education computing’, focuses on the technical aspects of language models, while the red cluster with concepts such as: ‘Teaching’, ‘curricula’, ‘contrastive learning’, ‘federated learning’ connotes innovative pedagogical methodologies, curriculum design and collaborative learning approaches mediated by generative artificial intelligence. Each cluster reflects a key theme within the conceptual framework of generative artificial intelligence in higher education, showing the interactions between the concepts and their relevance in the development of this field.

The analysis of clusters and their connections contributes to creating a holistic conceptual framework on generative artificial intelligence in higher education, integrating several key perspectives and dimensions that interact synergistically. Subsequently, a thematic analysis was performed in ATLAS.TI version 8, which allowed for a categorization process by grouping the codes into new emerging categories that constitute key processes for the sustainability of generative artificial intelligence in the development of transversal competencies in higher education (see Figure 4).

It is revealed that the sustainability of Generative Artificial Intelligence in higher education lies in the synergy between ethical appropriation, contextualized institutional management in technological infrastructure, teacher development, curricular restructuring, and pedagogical innovation. These five processes form an ecosystem where each element reinforces the others: without a solid ethical framework, pedagogical innovations can generate risks; without strategic institutional management, curricular transformation efforts lack support; and without trained teachers, generative tools will never reach their full educational potential. Only by coherently articulating these dimensions will we consolidate a Generative Artificial Intelligence integration model that is at once innovative, responsible, and enduring.

The critical and ethical appropriation of Generative AI requires higher education institutions to establish regulatory frameworks and spaces for reflection where both the capabilities and risks of these technologies are evaluated. This implies developing clear policies on data privacy, copyright, and algorithmic biases, as well as promoting critical literacy for students and faculty to understand how generative models are trained, operate, and mutate. Only through an institutional culture that recognizes the ambivalence of Generative Artificial Intelligence—its innovative potential and its threats to academic integrity—will it be possible to ensure its responsible and sustained use over time.

At the same time, institutional management of technological infrastructure requires aligning intelligent and contextualized investment in technological infrastructure—servers, cloud platforms, software licenses—with governance strategies that respond to the mission and characteristics of each university. This implies designing cross-cutting committees involving ICT, student welfare, research, and teaching areas, so that adoption decisions, monitoring of use, and updating of models are made in a contextualized, sustainable manner and with full knowledge of the financial, ethical, and social impacts.

Teacher development is vital to sustain Generative Artificial Intelligence in academia; teachers must have continuous training that combines technical aspects (model training and tuning, data governance) with pedagogical reflection on its applications and limitations. Mentoring programs, collaborative workshops, and communities of practice allow teachers to share experiences, co-create teaching resources, and keep abreast of a rapidly evolving field, ensuring that the integration of Generative Artificial Intelligence is coherent, innovative, and aligned with institutional goals.

Teacher development allows a curricular transformation that goes beyond the mere adoption of tools; it means rethinking learning objectives, evaluation methodologies, and content structure to take advantage of generative capabilities. Instead of treating Generative Artificial Intelligence as a complement, curricula should incorporate activities where language modeling and content generation become pedagogical axes, fostering human-machine co-creation competencies, critical thinking about automatically generated results, and adaptability to emerging information environments.

All this is the basis for driving pedagogical innovation powered by Generative Artificial Intelligence, which opens the door to unprecedented instructional designs: high-fidelity simulated laboratories, personalized tutorials with generative chatbots, and adaptive learning environments that respond in real time to the performance of each student. These methodologies promote an active role for the learner, facilitate safe experimentation, and allow for continuous formative assessments based on automated analysis of creative and cognitive processes, thus enriching the teaching-learning process.

RQ3. What is the distribution of the scientific production of generative artificial intelligence in higher education according to the type of document?

The distribution of scientific production on generative artificial intelligence in higher education shows a clear predominance of academic articles, followed by papers presented at conferences. This suggests that research in this field is mainly disseminated through peer-reviewed publications and specialized congresses, reflecting the relevance of the topic in the academic community. To a lesser extent, book chapters and reviews are found, indicating that there is also an effort to consolidate knowledge in more structured formats. Other types of documents, such as editorials, notes, books, letters, and errata, have little presence (see Figure 5).

As can be seen, the number of reviews, especially systematic ones, is considerably lower, which points to a crucial opportunity and need to strengthen this type of study to synthesize existing knowledge, identify trends, assess the impact of Artificial Intelligence in education, and guide future research with a solid foundation. This would also allow for the consolidation of scattered findings, provide more robust theoretical frameworks, and facilitate informed decision-making in curriculum design and AI literacy within higher education.

RQ4. How does the geographical distribution of publications behave?

The most relevant countries, according to the corresponding author, are shown below (see

Table 6. Most relevant countries, according to the corresponding author.

Country	Manuscripts	SCP *	MCP **	Freq ***	MCP_Ratio ****
China	56	5	7	0.021	0.583
United Kingdom	37	10	3	0.023	0.231
USA	21	13	4	0.03	0.235
Australia	19	8	10	0.032	0.556
Hong Kong	18	12	7	0.033	0.368
Spain	17	17	4	0.037	0.19

Notes—* SCP (Single Country Publications): Number of publications published exclusively by authors from a single country, i.e., without international collaboration; ** MCP (Multiple Country Publications): Number of publications published in collaboration with authors from other countries; *** Frequency: Number of articles published by corresponding authors from a country; **** MCP_Ratio: It is the ratio of international collaborative publications (MCP) to total articles. A higher value of MCP_Ratio indicates that a country has more international collaborations in proportion to its total articles.

).

The above is complemented by a graph that visually shows the proportion of publications from each country; the blue part of the bar represents publications by authors from a single country (SCP), while the red part represents publications in international collaboration (MCP). See Figure 6.

China leads in several articles, Hong Kong stands out for its high proportion of publications in collaboration with authors from several countries (international collaboration), while Spain and the USA prioritize research within their borders.

In raw data, the scientific production analysis reveals that China leads the ranking with 128 publications, followed by the United Kingdom with 107 and the United States with 95, consolidating its position as the three countries with the greatest contribution in the field. Australia occupies fourth place with 73 documents, followed by India with 59. In a second productivity band are Spain, with 43 publications; Indonesia and Jordan, with 41 each; finally, Malaysia and Peru, both had 38. In summary, scientific production is led by traditional powers, but the growing participation of emerging countries stands out, reflecting a greater geographic diversification of knowledge (see Figure 7).

RQ5. What thematic areas stand out in scientific production on the use of Generative Artificial Intelligence in higher education?

Below is a graph showing the distribution of thematic areas involved in the scientific production of generative artificial intelligence (AI) in higher education (see Figure 8).

The results show that Social Sciences stand out as the most represented field, suggesting a predominant interest in studying the impacts of generative AI from social, pedagogical, or educational perspectives. In second place, Engineering appears, with a notably lower but significant volume, indicating its importance in the technical development and application of generative AI. Other areas such as Mathematics, Arts and Humanities, and Health Professions present smaller but also relevant contributions related to technical methodologies, creative applications, or ethical implications in the context of AI in higher education, and finally, the other areas show a low representation, indicating that the connection between these disciplines and the use of generative AI in education is more limited. These results reflect the growing presence of an interdisciplinary perspective, where diverse areas of knowledge converge to explore and enhance the applications of generative AI in higher education.

RQ6. What are the most frequent terms associated with generative AI in higher education, and how has their frequency varied?

The figure shows how the academic community massively turned its attention in 2023–2024 to ChatGPT, large language models, and generative AI, which is reflected in a very pronounced spike in these keywords. From 2024 onwards, the growth slows down, indicating that the topic enters a consolidation stage. Meanwhile, more classical concepts of AI and contrastive learning, as well as learner research, continue to grow, but more gradually (see Figure 9).

4. Discussion

In compliance with Item 23a of the PRISMA guide, this section will be devoted to the general interpretation of the findings obtained in the context of other relevant evidence. In order not to be limited only to listing the identified transversal competencies, the discussion on how to generate them in a sustainable way using Generative Artificial Intelligence (GenAI) in higher education is expanded, thus providing a greater added value to this systematic review. The analysis of the included studies allows the identification of five strategic axes that, together, contribute to the sustainability of GenAI to develop transversal competencies in a lasting manner.

First, the critical and ethical appropriation of GenAI is revealed as an indispensable condition to avoid its uncritical use, guarantee academic integrity, and promote digital responsibility. This conclusion aligns with previous research that underlines the urgency of establishing solid ethical frameworks and digital literacy programs in artificial intelligence for the university environment. Thus, higher education institutions should foster critical digital literacy throughout the institution to ensure that both staff and students can use GenAI responsibly and ethically [63].

Secondly, institutional management of the technological infrastructure is seen as an enabling base. Available evidence shows that, without adequate and equitable technical support, GenAI integration tends to be restricted to fragmented experiences, limiting its systemic impact. Institutions should create comprehensive policies that address the ethical use of GenAI, ensuring transparency and accountability [64]. Building a robust technology infrastructure is critical to successful AI integration. This includes investing in cutting-edge technology and ensuring scalability to adapt to future growth [65,66]. Key challenges include resource constraints, data complexity, and the need for scalable solutions. In addition, institutions in developing regions may face unique challenges related to cost, expertise, and access to technology [67,68].

Third, teacher development for sustainable integration stands out as a key strategy for strengthening institutional capacities. The studies reviewed agree that teachers must be accompanied in their process of adaptation to the new tools, not only in the technical domain but also in their pedagogical, ethical, and reflective dimensions. Thus, Celik’s study [69] directly supports the teacher development axis for the sustainable integration of AI in higher education by showing that teachers require not only technological but also pedagogical and ethical knowledge to integrate AI tools effectively. His proposal for an Intelligent-TPACK framework reinforces the need for comprehensive teacher training that combines these skills, thus ensuring a reflective, ethical, and pedagogically relevant use of AI in educational contexts.

Woolf’s work [70] links to the axis of teacher development for sustainable AI integration by highlighting that the effectiveness of intelligent tutors depends not only on the software but also on teacher preparation. The author stresses that the lack of teacher training has been a key barrier to the adoption of these technologies in education. Thus, the development of teaching skills in the pedagogical and adaptive use of intelligent systems becomes essential to take full advantage of their transformative potential in higher education. In this regard, one study points out that, although AI is increasingly applied in educational processes, many research studies lack a clear pedagogical basis, which underlines the importance of training teachers not only in technical aspects but also in formative and ethical approaches for meaningful and sustainable integration [71].

Fourth, curriculum transformation mediated by GenAI is a fundamental area for renewing content, methodological approaches, and evaluation criteria, responding to new learning scenarios. This curricular transformation requires a profound revision of the educational design, as suggested by innovative experiences with emerging technologies in university contexts [72].

Finally, pedagogical innovation with GenAI makes it possible to dynamize the teaching-learning processes through more active, personalized, and collaborative proposals. This approach is in line with studies that document the potential of GenAI to foster adaptive, meaningful, and student-centered learning. In this regard, one paper highlights pedagogical innovation as a driver for transforming higher education through the use of Generative Artificial Intelligence (GenAI), while positioning GenAI as a catalyst for pedagogical change [73].

Taken together, these results indicate that the sustainability of GenAI in higher education cannot be approached from a unidimensional perspective. On the contrary, it requires an articulated strategy that integrates technological infrastructure, teacher training, curriculum redesign, pedagogical innovation, and critical appropriation of technology. This systemic view provides relevant evidence for the design of educational policies, institutional decision-making, and future lines of research.

A highly cited study shows low teacher involvement and weak connection with pedagogical frameworks in AI applications in higher education [74]. Their findings support the need for a sustainable integration of GenAI based on teacher development, ethical appropriation, and pedagogical innovation, as proposed by this systematic review.

Another study provides a key perspective by highlighting student perceptions of GenAI in higher education. While students recognize benefits such as personalized learning and support in writing and research, they also express concerns about accuracy, ethics, and the impact on their personal development. [75]. These findings reinforce the need to integrate the student voice in the design of sustainable strategies that promote transversal competencies, ensuring a pedagogically meaningful and ethically responsible implementation of GenAI.

The UTAUT model proposed by Venkatesh et al. offers a robust framework for understanding the factors that determine the acceptance and use of new technologies, such as GenAI, in educational contexts [76]. Its incorporation makes it possible to analyze how variables such as performance expectation, effort required, social influence, and facilitating conditions can influence the sustainable adoption of GenAI by faculty and students. This perspective is key to designing institutional strategies that promote transversal competencies, anticipate barriers, and facilitate pedagogically meaningful integration.

Mishra and Koehler’s TPCK framework provides a key theoretical basis for understanding the sustainable integration of GenAI in higher education [77]. Its emphasis on the complex interaction between technology, pedagogy, and content is aligned with the five strategic axes identified for the responsible and effective use of GenAI, especially in teacher training and curricular transformation. Thus, TPCK supports the need for comprehensive professional development that fosters a critical, ethical, and contextualized appropriation of these technologies, ensuring their lasting impact on the development of transversal competencies.

The article by Dwivedi et al. [2] provides a multidisciplinary perspective that complements the analysis of the sustainable integration of GenAI in higher education for the development of transversal competencies. Its reflections on opportunities, ethical and legal challenges, and the need for transparency and capacity building coincide with the strategic axes identified, especially in terms of critical appropriation, teacher training, and institutional management. In addition, the call to deepen the understanding of biases and the optimal combination between humans and GenAI reinforces the importance of a reflective and contextualized pedagogical approach, as proposed by the TPCK framework, to ensure responsible and effective use of these technologies in the formation of transversal competencies. To address these challenges, it is essential to establish ethical guidelines and policies aligned with the principles of academic integrity [78].

5. Conclusions

Remarkably, there are no results before 2023 with the established search equation; this is evidence of the recent explosion of interest in generative AI, especially with the launch and popularization of technologies such as Large Language Models and AI tools such as ChatGPT. For its part, the notable exponential increase in 2024 reflects how generative AI has established itself as a key topic of interest in higher education, although studies with the descriptors “UTAUT2 Model” OR “Deep Learning” have been found to have literature associated with higher education before 2023. To build a comprehensive framework and guide future research towards a more structured and global view, it is essential to address the key issues addressed in the research questions.

In compliance with Item 23b of the PRISMA statement, it is necessary to recognize that the evidence included in this review on the use of GenAI for the development of transversal competencies in higher education has limitations that affect its validity and generalizability. First, most of the studies identified focus on very specific tools, such as ChatGPT, without exploring the full spectrum of emerging generative technologies, which reduces the ability to extrapolate findings to other similar environments or applications. In addition, most of the research reviewed adopts a descriptive approach and lacks longitudinal or robust comparative designs to objectively evaluate the effectiveness of GenAI in improving competencies such as critical thinking, creativity, or problem-solving. Likewise, there is a geographical bias towards English-speaking contexts, while regions with less prominence in traditional databases, where relevant innovative practices could exist, are underrepresented. These shortcomings limit not only the diversity of evidence but also the strength of the conclusions that can be reached in terms of international applicability and sustainability of the proposals.

In relation to Item 23c, the limitations of the review processes used are present in the evaluation of the databases and the selection criteria applied. Despite having carried out systematic searches in Scopus, Web of Science, and PubMed, the omission of repositories specialized in educational sciences (e.g., ERIC) and computer engineering (such as IEEE Xplore) causes relevant publications that directly or indirectly address the integration of GenAI in academic contexts to be left unreviewed. In addition, the language restriction to articles in English and Spanish introduces a linguistic bias that excludes studies in other languages, potentially biasing the overall view of existing practices and innovations.

By Item 23d, the implications of the results obtained for educational practice, institutional policies, and future lines of research are particularly relevant. In the practical sphere, the urgency of designing teacher training programs that comprehensively articulate technological, pedagogical, and ethical knowledge so that teachers acquire competencies to implement GenAI critically and effectively in the classroom is confirmed. This includes technical training and the development of skills to evaluate the quality of AI-generated content and the ability to integrate these tools into the curriculum with clear formative purposes. From the policy point of view, it is essential to establish institutional regulatory frameworks that promote collaborative governance (involving students, teachers, IT staff, and academic authorities) and guarantee sustainable investments in technological infrastructure to ensure equitable access and avoid digital divides. Likewise, digital literacy programs should be promoted to facilitate the ethical appropriation of GenAI by the entire university community.

As for future research, the need for empirical and longitudinal undertakings that rigorously measure the impact of various Generative AI solutions on transversal competencies is highlighted, as well as studies that include sources of grey literature and local repositories to enrich the picture with less visible contextual experiences. It is also recommended that mixed and multidisciplinary approaches be adopted to understand not only the quantitative results but also the qualitative dynamics behind the integration of Generative AI, including student and teacher perceptions. Similarly, it is relevant to evaluate which governance models and professional development strategies are most effective according to variables such as organizational culture, available resources, and teacher profiles. The development of standardized evaluation frameworks will facilitate the systematic measurement of transversal competencies, favoring the incorporation of these indicators in accreditation and continuous curricular improvement processes.

The systematic review demonstrates that the sustained development of transversal competencies—academic integrity, critical thinking, ethics, innovation, creativity, communication, responsibility, collaboration, autonomous learning, digital literacy, self-regulation, leadership, AI literacy, AI ethics, and flexibility—is achieved only when five strategic pillars converge: a critical and ethical appropriation of GenAI that safeguards academic integrity, stimulates critical thinking, and promotes ethical responsibility; institutional management of technological infrastructure that ensures equitable access and enhances both digital literacy and AI literacy; robust faculty development that empowers educators to integrate AI creatively and innovatively, thereby reinforcing their capacity for creativity, communication, and leadership; a GenAI-mediated curricular transformation that aligns content and pedagogical approaches to foster autonomy, self-regulation, and collaboration; and, finally, pedagogical innovation with GenAI that provides flexible, personalized learning environments, encouraging both experimentation and active, collaborative learning. Only by integrating these pillars can an educational ecosystem be established in which GenAI sustainably drives the comprehensive formation of professionals equipped to meet the challenges of the knowledge society.

It is concluded that the sustainable integration of generative artificial intelligence (GenAI) in higher education requires an articulated institutional strategy that combines regulatory guidelines, pedagogical innovation, and faculty professional development. First, it is recommended that institutions incorporate transversal competencies such as critical thinking, creativity, ethics, communication, and self-regulation explicitly into curricula through activities that promote the critical use of GenAI tools. For example, assigning tasks where students use ChatGPT to co-construct texts that must then be analyzed, corrected, and argued from an ethical and reflective perspective allows them to simultaneously develop cognitive, metacognitive, and socioemotional skills.

It is also necessary to establish clear regulatory frameworks on the ethical and responsible use of these technologies, including protocols against fraudulent use and practical guidelines for their pedagogical integration. These frameworks should be constructed in a participatory manner and incorporated into academic regulations. At the same time, it is essential to implement continuous teacher training programs that address not only technical aspects of use but also their pedagogical, ethical, and evaluative dimensions. For example, practical workshops where teachers design and validate rubrics to evaluate tasks generated in part with GenAI support allow progress towards informed and critical adoption.

In the same vein, it is recommended that interdisciplinary pedagogical-technological support units be created to accompany teaching teams in the incorporation of these tools, offering advice, resources, and models of good practice. At the student level, it is essential to promote training spaces in digital ethics, artificial intelligence literacy, and critical thinking through specific modules, experimentation laboratories, or guided activities in which outputs generated by language models are analyzed.

Finally, the need for inclusive policies for access to technological infrastructure and connectivity is highlighted, as well as the development of indicators and evaluation instruments to measure the impact of the use of GenAI in the development of transversal competencies. Only through this coherent integration of policies, training, infrastructure, and critical institutional culture will it be possible to consolidate an educational ecosystem where GenAI enhances the comprehensive and adaptive training of students in the digital era.

Registration and protocol: In accordance with PRISMA Item 24a, this review was not registered in any database. Consequently, no review protocol was prepared or made publicly available (Item 24b), and there are no amendments to report (Item 24c).