Abstract
Generative AI is revolutionizing education by enhancing personalized learning, fostering innovation, and transforming traditional teaching methodologies, making it a critical tool for the future of education. This study aims to explore the impact of generative AI (Gen-AI) technologies, focusing on academic and learning performance, publication trends, and thematic developments through a comprehensive bibliometric analysis. This study employs a bibliometric analysis using data from WoS and Scopus, focusing on publications from 2015 to 2025. The analysis includes keyword co-occurrence, bibliographic coupling, and citation analysis to map the evolving landscape of Gen-AI adoption by students and teachers in education. The study reveals a significant surge in publications on generative AI in education, highlighting key themes such as ChatGPT, and higher education. The United States, Australia, and United Kingdom lead in research contributions, with diverse topics explored across major journals. This study provides a novel bibliometric analysis of generative AI adoption on academic performance, offering unique insights into publication trends, influential journals, and emerging themes, thereby contributing to the understanding of generative AI’s evolving role in education. The results highlight the need for further longitudinal studies to explore the long-term impact of generative AI on educational practices.
Plain language summary
Generative AI (Gen-AI) is changing how we learn and teach. It helps personalize learning, encourages creativity, and improves traditional learning and teaching methods. This study looks at how adopting Gen-AI impacts education, focusing on academic and learning performance, publication trends, and key themes in recent research. Using data from major research databases (Web of Science and Scopus) for the years 2015 to 2025, the study maps out how Gen-AI is being discussed in academic circles. This study analyzes the connections between keywords, influential publications, and citations. Findings show a sharp increase in studies about Gen-AI in education. Key topics include its effects on student-teacher performance and the role of technology in classrooms. The United States, Australia, and the United Kingdom are leading research efforts, with studies published in top journals exploring various issues. This study offers a fresh perspective on how Gen-AI shapes education, highlighting trends, important journals, and emerging focus areas. It provides valuable insights into how this technology is transforming learning and teaching around the world.
Keywords
Introduction
The remarkable advancement of artificial intelligence (AI) technology in the 21st century has significantly altered many industries, including education (Biagini, 2025). The transformational potential of generative AI (Gen-AI) models that can produce text, images, and other media makes them stand out among the new AI technologies (Kanbach et al., 2024). Chatbots, automated content production systems, virtual teaching assistants, and personalized learning platforms are just a few of the technologies and applications that fall under the umbrella of generative artificial intelligence. These resources are made to help teachers and students by offering scalable, personalized, and interactive learning opportunities (Bahroun et al., 2023). The rapid advancement of Gen-AI technologies, such as OpenAI’s ChatGPT, DeepSeek, DALL-E, and Meta’s LLaMA2, has caused a significant shift in education (Lim et al., 2023; Peng et al., 2025; Radanliev, 2024). Academic attainment is greatly impacted by these Gen-AI models’ capacity to mimic human material through deep learning approaches. Because these technologies can produce creative artwork, well-written essays, and even code, they are growing in popularity among educators and students (Shahzad, Xu, An, Asif, & Javed, 2025).
According to a previous report by Google Cloud (2023), Gen-AI creates new text, photos, audio, video, and music by using machine learning models to find patterns in preexisting human-generated content. On the other hand, data prediction based on past behavior was the primary focus of earlier AI systems. The market for Gen-AI in education is anticipated to expand rapidly, with a compound annual growth rate (CAGR) of 39.5% between 2024 and 2033 (Precendence Research, 2023). This would bring the market’s value from USD 299.8 million in 2023 to over USD 7,701.9 million by 2033. This increase implies that more people are realizing the benefits of Gen-AI in education (Bandi et al., 2023). Due to technological developments and a rise in consumer demand for e-learning solutions, the e-learning industry, which was valued at USD 399.3 billion in 2022, is also expected to grow at a compound annual growth rate (CAGR) of 14% between 2023 and 2032 (Global Generative AI In Edtech Market, 2024).
For students and teachers today, implementing Gen-AI in the classroom is becoming more and more crucial (Ivanov et al., 2024). Gen-AI technology deployment in the educational field holds promise for enhancing student performance, providing tailored assistance, and stimulating creativity (Kamalov et al., 2023). By enabling collaborative learning settings, generating adaptive learning paths, and offering immediate feedback, Gen-AI systems can improve learning for pupils (Gilson et al., 2023). These tools can help teachers create instructional materials, automate administrative activities, and provide insights into students’ learning styles (Z. Gao et al., 2024). However, there are questions over the effectiveness, ethical ramifications, and long-term repercussions of using Gen-AI due to its varied and complicated effects on teachers’ and students’ academic performance (Bahroun et al., 2023).
These Gen-AI technologies, such as ChatGPT, DALL-E, and LLaMA2, are of importance to the education industry due to their rapid advancement (Radanliev, 2024). These Gen-AI tools support educators and learners on an individual basis, foster creativity, and enhance educational opportunities in new ways. DALL-E and LLaMA2 offer unique opportunities for content creation and visual learning (Lim et al., 2023), ChatGPT, a language-based Gen-AI, can assist with writing, brainstorming, and teaching (Ivanov et al., 2024). Similar to this, DeepSeek-R1, which was released on January 27, 2025, attracted a lot of users and swiftly rose to the top of the generative AI app download charts (Peng et al., 2025). By providing individualized learning support, expediting research, and boosting academic output, DeepSeek also improves educational success (Cui, 2025). Understanding how these Gen-AI tools affect academic performance is crucial given their growing integration into the education sector (Shoaib et al., 2024). The impact of Gen-AI technologies on students’ and instructors’ writing, creativity, problem-solving, and academic performance is the main emphasis of this study. This study also discusses ethical difficulties in education, the potential replacement of traditional teaching and learning methods with Gen-AI-generated content, and other related considerations (Shahzad, Xu, An, Asif, & Javed, 2025).
Despite the attainable advantages, nothing is known about how the use of Gen-AI affects academic achievement in real-world scenarios. The technological aspects of AI in education or its effects on pedagogical practices have been the main focus of previous studies (Bahroun et al., 2023; Hashmi & Bal, 2024; Shoaib et al., 2024), with little attention paid to thorough bibliometric analyses that look at the scope and depth of research in this area. This study employs a bibliometric approach, which entails a qualitative examination of scholarly literature to provide insightful information about the main themes, developing trends, and unmet research needs in the use of generative AI in education. By performing a bibliometric analysis of the use of Gen-AI in education, including ChatGPT, DeepSeek, DALL-E, Gemini, Microsoft Copilot, and LLaMA2, and its effects on students’ and teachers’ academic achievement, this study seeks to close this gap. This study aims to identify the major themes, influential authors, widely used approaches, and new developments in the discipline by carefully examining the body of existing literature. The following research questions are the focus of the current investigation.
This study’s design is as follows: In Section 2, the current research literature is discussed. Section 3 outlines the methodology of the bibliometric analysis as well as the techniques for data collection and processing. Section 4, presents the findings, which also contains a list of nations, journals, significant publications, themes, and research trends. Finally, the study’s discussion, consequences, and future directions are covered in Section 5.
Literature Review
These days, generative AI has the potential to totally transform a variety of inclusive teaching methods (Biagini, 2025). Gen-AI has garnered a lot of interest in teaching and learning due to its potential to alter pedagogical techniques and improve learning opportunities in a variety of academic fields, such as engineering, mathematics, and the social sciences (Bahroun et al., 2023). International organizations like UNESCO and individual educational institutions are debating the best approaches to integrate and use Gen-AI in the classroom in order to handle the significant influence of this technology on teaching and learning (Crawford et al., 2023; Fullan et al., 2023). Additionally, studies have demonstrated that these technologies can enhance instructor experiences, boost student engagement, and personalize learning, all of which can result in better outcomes (Farrukh Shahzad et al., 2025; Ivanov et al., 2024). A prior study Shoaib et al. (2024) suggests that Gen-AI-driven, personalized teaching systems that adjust knowledge to each student’s learning style and pace may improve content retention and comprehension. In higher education, Gen-AI fosters critical thinking and the capacity for innovative problem-solving (Z. Gao et al., 2024).
ChatGPT, created by OpenAI, became live in November 2022 and, despite breaking the previous record for the user application with the fastest expansion, gained 1 million subscribers in just 5 days and surpassed 100 million users in 2 months (Kanbach et al., 2024). With 175 billion parameters, GPT-3 represented a significant breakthrough in language modeling and served as the basis for ChatGPT, which rapidly became well-liked in industries like healthcare and education (Wang et al., 2023). Released on March 14, 2024, the newer GPT-4 model has a substantial increase in processing capability compared to its predecessor, with 170 trillion parameters (Raiaan et al., 2024; Yenduri et al., 2024). Currently, another Gen-AI technology, DeepSeek, a Chinese AI firm, supports education through its advanced language models by offering tools that enhance learning and teaching efficiency. Hangzhou DeepSeek Artificial Intelligence Basic Technology Research is a Chinese firm focused on creating large language models (LLMs) (Peng et al., 2025). It attracted widespread recognition following the launch of its DeepSeek-R1 model in January 2025 (Franzoni et al., 2024; Kumar et al., 2025).
According to a prior study, students and teachers acquire these critical competencies more successfully when Gen-AI technologies that create problem scenarios and offer iterative feedback are used than when traditional learning methodologies are used (Shahzad, Xu, An, Asif, & Javed, 2025). Furthermore, Gen-AI technology can help teachers by reducing some of their administrative responsibilities so they can concentrate more on instruction and less on assessment and grading (Kanbach et al., 2024). Through realistic conversational scenarios, text-generative language models such as ChatGPT-4 and LLAMA2 can provide learners with a personalized and dynamically adaptive platform to improve their language proficiency in authentic situations (Abd-Alrazaq et al., 2023; Radanliev, 2024), and students in the education sector make extensive use of Gen-AI to find and learn new skills and competencies that improve students’ cognitive abilities and learning capacities (Budhathoki et al., 2024).
Engagement and communication amongst interactive conversational agents like Gen-AI technologies have increased the quality of training (Bahroun et al., 2023). However, prior research has also shown that DALL-E improves comprehension and academic performance (Hashmi & Bal, 2024). Research on ChatGPT, Gemini, and Microsoft Copilot use in higher education has been scant (Kurtz et al., 2024; Shahzad, Xu, An, Asif, & Javed, 2025). Similarly, Gen-AI tools like DeepSeek, LLAMA2, DALL-E, and ChatGPT could be beneficial tools for students’ individual research, helping them pass their classes and become better writers (Peng et al., 2025; Radanliev, 2024). These Gen-AI tools are useful in the education sector for brainstorming, data synthesis, and data identification. In order to motivate students and faculty to use critical and creative thinking to address real-world problems, educators should incorporate Gen-AI into their curricula (Shahzad, Xu, An, Zahid, & Asif, 2025; Shoaib et al., 2024). To strengthen this study’s arguments, it is essential to address the ethical challenges and AI governance associated with generative AI, as past studies highlighted (Al-kfairy et al., 2024; Perkins, 2023). Incorporating AI ethics and governance frameworks can provide a more comprehensive understanding of the responsible incorporation of Gen-AI technologies in education (Fosso Wamba et al., 2025).
Gen-AI boosts efficacy, efficiency, and productivity, which greatly enhances education. To remain competitive with these technologies, the education sector—particularly in universities—must constantly innovate (Shahzad, Xu, An, Asif, & Javed, 2025). Teachers and students should lead the adoption of Gen-AI to ensure its safe and effective use in education (Budhathoki et al., 2024; Shahzad, Xu, An, Zahid, & Asif, 2025). Its adaptability and capacity to impart extensive knowledge have quickly made it popular among students in both developed and developing nations (Cotton et al., 2024). The widespread use of these Gen-AI technologies raises questions about educational integrity, over-reliance on technology, and the potential for compromised critical thinking skills (Chan, 2023). The ability of students to generate high-quality information with less effort may lead to a decline in traditional study habits and a reduction in the depth of their understanding of important subjects (Lo et al., 2024). By employing Gen-AI to assist with curriculum design, grading automation, and student performance analytics, educators can employ more focused and effective teaching strategies (Budhathoki et al., 2024). Nonetheless, there are challenges and worries regarding the effectiveness and ethical consequences of using these tools in the classroom (Perkins, 2023). Therefore, to successfully navigate the challenges of Gen-AI in education, the government or institutions should establish guidelines or promote digital literacy.
Methodology
Our study’s approach is based on the mapping that is based on (Aria & Cuccurullo, 2017), which (Asif et al., 2024; Hashmi & Bal, 2024) further expanded upon. This method consists of five phases: design of the study, gathering, analyzing, displaying, and interpreting data. The phases are based on (Donthu et al., 2021; Lodhi et al., 2024). The methods and approaches used in the current investigation are displayed in Figure 1.
PRISMA flowchart, inclusion, and exclusion process for the bibliometric analysis study.
Research Design (Phase 1)
Topic of the Study and Objectives
This section outlines the central theme of the research paper, which focuses on exploring how the adoption of Gen-AI can strengthen the academic performance of both students and teachers. The study aims to investigate this topic through a bibliometric analysis, mapping out existing research, identifying trends, and highlighting key areas of interest within this field.
Selection of Techniques
Scientometrics investigates the tools, developments, and structures that are utilized in the building of new theories and techniques. A subfield of scientometrics called bibliometrics studies how a given scientific subject behaves and evaluates scientific literature using scientific methodologies (Lim & Kumar, 2024). The science of science was established in the latter stages of the 20th century (H. Gao & Ding, 2022). It is therefore being used increasingly frequently to assess scientific research. Furthermore, data from various scientific disciplines can be more easily evaluated thanks to bibliometric analysis (Ellegaard & Wallin, 2015). It offers fascinating information about the relationships between research and creativity. In this sense, the development of science and the accessibility of a multitude of databases to researchers have led to the growth of the discipline of bibliometrics. Senior management must use bibliometrics as a critical tool when deciding what new research projects to pursue. The understanding and justification of the patterns of work and scientific output are aided by bibliometric research (Bahroun et al., 2023). In bibliometric analysis, indicators are employed to illustrate the progress made in science. Subject areas are used initially to monitor the outcomes of scientific activity, then activity indicators to obtain data regarding the extent and significance of research endeavors, followed by association indicators to assess researcher interactions. This study used methods including keyword co-occurrence analysis and descriptive statistical analysis to address the research themes (Shahzad, Asif, et al., 2025). Descriptive statistical evaluation allows for the recognition of the relevant affiliations, authors, and different countries (Franco et al., 2024).
Selection of Tools
We used “VOSviewer” (Guo et al., 2019) and “Biblioshiny” (Shahzad, Asif, et al., 2025) with RStudio for the data analysis, which is the third part of the research technique (Asif et al., 2024). Biblioshiny is used for descriptive analysis, whereas VOSviewer is used for network analysis in the program. Non-programmers can use Biblioshiny to perform bibliometric analysis (Ejaz et al., 2022). It can do general statistical calculations and create a three-field plot to show the affiliation between particular items (Agbo et al., 2021). Scholars may thoroughly process and analyze bibliographic data with the Biblioshiny tool, and then export the information into a tabular and graphical format for additional research and discussion (Bahoo et al., 2023; Lim & Kumar, 2024; Singh & Bashar, 2023). In contrast, VOSviewer allows the creation of network-based country maps, keyword maps derived from connected networks, and maps with a multitude of components. Data mining, grouping, mapping, and database article gathering are all possible with the VOSviewer application (Shahzad, Asif, et al., 2025).
Selection of Keywords
This study extracted data from the WoS and Scopus databases using the following keywords: ((“Generative artificial intelligence” OR “Adoption of Gen-AI technologies” OR “adoption of generative AI” OR “Generative AI” OR “ChatGPT” OR “DALL-E” OR “LLaMA2” OR “Gemini” OR “DeepSeek” OR “Microsoft Copilot”) AND (“Academic performance” OR “Learning performance” OR “Students” OR “Teachers” OR “Technology” OR “Educational technologies” OR “educational transformation” OR “research trends,” OR “ethical challenges”)). These keywords were paired with terms like “Academic performance,”“Learning performance,”“Students,”“Teachers,” and “Education sector” to comprehensively capture studies linked to the adoption of Gen-AI technologies in education. This approach ensured the identification of relevant literature focusing on how Gen-AI tools like ChatGPT impact the academic performance of students and teachers.
Selection of Databases
All relevant publications for this investigation were obtained from reputable databases for academia, such as WoS and Scopus. The purpose of this data gathering was to produce a comprehensive and trustworthy dataset that reflects the most recent academic advancements in our field of study. Most bibliometric analyses employ a single data source (Lodhi et al., 2024). Scopus is open from Elsevier, and WoS is presented by Thomson Reuters (Mongeon & Paul-Hus, 2016). However, we selected both databases for our bibliometric analysis. WoS was selected for two reasons: 1) it is one of the largest databases available, containing the best works in a variety of fields of academia between 1900 and the present, making certain the excellent caliber of the items in the index (Bahroun et al., 2023); 2) the database serves as the primary information source for those who conduct literature reviews (H. Gao & Ding, 2022; Lim & Kumar, 2024). Conversely, Scopus offers several advantages, such as: (1) being the largest gathering of peer-reviewed literature; (2) reducing the possibility that information would be lost in searches; (3) being easy to access; (4) provide tools for analyzing and visualizing data; (5) enabling the download of samples in multiple file types; and (6) providing various types of data (Baas et al., 2020; Brzezinski, 2015; Singh & Bashar, 2023).
Data Collection (Phase 2)
Extracting and Cleaning Data
The data extraction process for this research involved a comprehensive search in both the WoS and Scopus databases, focusing on the title, abstract, author keywords, and keywords plus sections. The data was extracted on May 01, 2025, covering documents from 2015 to 2025. Initially, 5,872 documents were identified in WoS. Three inclusion/exclusion criteria were applied to refine the selection. First, only documents categorized under Education, Educational Research, and Management were retained, reducing the number to 1,873. Second, early access articles, proceeding papers, review articles, and editorial materials etc., were excluded, further narrowing the selection to 1,122 documents. Finally, non-English articles were excluded, resulting in a final set of 1,063 articles suitable for bibliometric analysis (see Figure 1).
Similarly, in Scopus, 12,506 documents were initially retrieved between 2015 to 2025. After applying the first criterion, which limited the selection to the Social Sciences subject area, 5,221 documents remained. The second criterion excluded conference papers, reviews, book chapters, notes, conference reviews, books, editorials, letters, errata, etc., leaving 3,537 documents. The third criterion excluded non-English articles in languages, resulting in a final set of 3,238 articles. These filtered documents from both databases were then used for the bibliometric analysis, ensuring a focused and relevant literature base for the study (Lim & Kumar, 2024).
Removing Duplicates
After retrieving the final sets of articles from WoS (1,063) and Scopus (3,238), we utilized RStudio to merge the datasets from both databases. During this process, 898 duplicate entries were identified and removed. This careful consolidation resulted in a final collection of 3,403 unique articles, which were then used for the bibliometric analysis. This step ensured that the analysis was based on a comprehensive and non-redundant dataset, strengthening the accuracy and reliability of the research findings.
Analysis (Phase 3)
Descriptive
A statistical method called bibliometric analysis offers a quantitative study of material that has been published in academic journals (Ramana et al., 2024). By employing bibliometric analysis, the most prominent and productive journals, authors, studies, countries, and universities are ranked (Ellegaard & Wallin, 2015). It is often used to visually depict literary works, such as words, journals, articles, and authors, in both professional and educational contexts (Bahroun et al., 2023).
Network Analysis
Network analysis is a method that looks at the relationships among words in certain texts (Bornmann et al., 2018). To display the co-occurrence network of keywords, VOSviewer was used (Singh & Bashar, 2023), which consists of using network visualization to build a map-based word network representing the themes (Bahroun et al., 2023; Ellegaard & Wallin, 2015; Lodhi et al., 2024). Both Biblioshiny and VOSviewer are helpful programs for producing visual representations of the literature. Graphs within the scientific community can display hotspots, new trends, and complex networks (Ejaz et al., 2022; Lim & Kumar, 2024). In this study, we looked at the co-occurrences of the keywords that different authors used to identify popular subjects in the chosen articles.
Display (Phase 4)
This phase focuses on summarizing important data regarding the number of publications released within the scientific community each year and the nations and journals that publish the most articles. This calls for the visualization and summarizing of data, including the number of scholarly articles published each year, the recognition of authors who have had the biggest impact on the field, the identification of the countries where this type of research occurs most frequently, and the presentation of the journals that publish a significant number of relevant contents. Key themes that were extracted from the papers are also covered at this point. These conversations can provide important new perspectives on the recurrent patterns and issues affecting the scholarly discussion in a particular field.
Interpretation (Phase 5)
Phase 5, also referred to as the phase of interpretation which is crucial for data analysis initiatives. At this point, researchers examine their findings to understand the trends and patterns the data reveal. It comprises a careful analysis of these trends, considering the study’s objectives and the corpus of previous research. By contextualizing the data and considering real-world implications, researchers attempt to arrive at well-founded conclusions that contribute to a set of information in their respective domains. At this point, the data’s complete significance becomes clear, and scholars gain a deeper comprehension of their topic.
Findings
The next sections, which each respond to one of our five research questions, offer the findings of the bibliometric analysis.
Descriptive Analysis
Response to RQ1
Figure 2 shows a dramatic surge in the number of articles published over the years, particularly from 2021 onward. Between 2015 and 2020, the output remained extremely low or stagnant, with zero publications recorded in six of those years. A modest increase occurred in 2021 with five articles, followed by a slight dip in 2022. However, 2023 marked a turning point with a significant rise to 446 articles, and this growth intensified in 2024, reaching a peak of 1,871 publications. Although there is a decline in 2025 to 1,076 articles, the overall trend indicates a rapidly growing interest and scholarly attention in the topic, likely driven by recent technological advancements or emerging relevance in the field.
Annual scientific production.
Response to RQ2
The analysis of publication sources indicates a strong concentration of research on the adoption of generative Gen-AI in education within a few prominent academic journals. “Education and Information Technologies” leads substantially with 155 articles, establishing itself as the most influential platform in this domain (Figure 3). It is followed by “Computers and Education: Artificial Intelligence” with 100 articles, and “Education Sciences” with 77, reflecting their central role in shaping scholarly discourse on Gen-AI in educational contexts. Other key journals include “JMIR Medical Education” (49), “Sustainability (Switzerland)” (43), and both the “International Journal of Human-Computer Interaction” and “Journal of Applied Learning and Teaching” with 42 articles each. “Frontiers in Education” and “Interactive Learning Environments” each contribute 41 articles, while the “Journal of Chemical Education” follows with 37. This distribution reveals a multidisciplinary interest in the topic, with both general education technology and domain-specific journals emerging as significant outlets for research dissemination.
Most productive journals.
Table 1 showcases the most cited scholarly articles on the adoption of Gen-AI, particularly ChatGPT, in education, reflecting the surge of academic interest in this transformative technology. The top-cited article, “Chatting and cheating: Ensuring academic integrity in the era of ChatGPT” by (Cotton et al., 2024), has amassed 2,199 citations, emphasizing the urgent concern over academic honesty in AI-integrated learning environments. Closely following is Rudolph et al. (2023a) with 2,134 citations for their provocative article questioning the future of traditional assessments in light of ChatGPT’s capabilities. Gilson et al. (2023) further extend the discussion into medical education, with 1,687 citations for their analysis of ChatGPT’s performance on the USMLE. Articles by Chan (2023) and Adiguzel et al. (2023) focus on AI policy frameworks and ChatGPT’s transformative educational potential, indicating a wide spectrum of scholarly attention. The presence of journals such as Journal of Applied Learning and Teaching, JMIR Medical Education, and Sustainability demonstrates that discourse on Gen-AI spans across multiple educational subfields—from ethics and policy to practical classroom implications. Overall, the high citation counts underscore the prominence and impact of these works in shaping contemporary conversations about AI’s role in reshaping pedagogical practices and educational leadership in the digital era.
Most Cited Articles.
Network Analysis
Response to RQ3
Table 2 illustrates the bibliographic coupling analysis, which reveals the most productive countries in the field of generative AI adoption in education, highlighting the global distribution of research activity. The United States leads significantly with 136 documents and 991 citations, demonstrating its central role in this research area. It also shows a high total link strength (65,504), indicating strong connections and influence within the global research network (see Figure 4). However, its average citations per document (7.29) suggest that while the U.S. produces the most research, the impact of each article is moderate. With 42 papers, Australia stands out for its high average citations per document (16.52), indicating a substantial effect on its research outputs despite having fewer publications than the U.S. The U.K., with 30 papers and an average of 21.97 citations per article, also exhibits a significant influence, reinforcing its strong research presence in this domain.
Most Productive Countries Based on Bibliographic Coupling.
Most productive countries based on bibliographic coupling.
Other countries like China, the United Arab Emirates, and Saudi Arabia contribute notably to the field, with varied average citation impacts. For instance, China has a lower average citation rate (3.87), while the United Arab Emirates has a higher average (11.05), reflecting differing levels of research influence across these nations. Notably, Singapore, despite contributing only 12 documents, has an exceptionally high average citation rate of 64.75, suggesting that its research outputs are among the most influential globally. In contrast, countries like India, with 19 documents but only 0.79 average citations, show a significant volume of research with comparatively low impact. Overall, this analysis underscores the leadership of countries like the United States, Australia, and the United Kingdom in driving the global research agenda on Gen-AI in education. At the same time, it highlights the emerging contributions from other regions, indicating a growing international interest and diversification in this research field.
Response to RQ4
Figure 5 illustrates the relationships between the three elements—countries, authors, and keywords—depicted in the three fields plot. Gray connections indicate the relationships between the components of the three-field plot. This link starts with the author’s national origin and continues with the author’s relationship to the keywords. The sizes of each rectangle in the list indicate how many papers are related to that particular component. In this layout, the author’s nation is on the left. The United States of America, which is home to several writers, has produced the most publications about Gen-AI in education. The second element in this three-field plot is the author’s name, which appears in the middle of the plot. These authors produce the most works. Common keywords are displayed and linked to each author on the right side of the three-field plot. The plot has a list of frequent keywords such as “ChatGPT,”“artificial intelligence,”“generative AI,”“higher education,”“large language model,”“academic integrity,”“chatbots,”“education,” and “AI”.
Keyword, author, and country interrelations.
Response to RQ5
The co-occurrence of keyword analysis reveals the prominent themes and topics in the current research landscape of generative AI adoption in education. By setting a threshold of 10 occurrences, we identified 62 significant keywords out of a total of 2,831, indicating a focused set of recurring concepts within the literature. These keywords were then grouped into three distinct clusters (Cluster 1 in Red, Cluster 2 in Green, and Cluster 3 in Blue in Figure 6 and Table 3), each representing a thematic area within the field. These clusters illustrate the diverse yet interconnected themes that are currently shaping research on Gen-AI in education.
Network analysis of co-occurrence of keyword analysis.
Cluster Analysis of Co-Occurrence of Keyword Analysis.
Note. OCR = occurrences.
Cluster 1 focuses on “Generative AI’s impact on higher education and academic integrity,” which is a central concern in the current educational landscape, especially in the context of this study. Higher education’s quick embrace of Gen-AI tools like ChatGPT encouraged both enthusiasm and apprehension among educators and researchers. On one hand, these technologies offer unprecedented opportunities to enhance learning, foster creativity, and personalize educational experiences. However, they provide serious obstacles to upholding integrity in academia, as they can be misused for plagiarism, cheating, and other forms of academic dishonesty.
This study investigates how Gen-AI is reshaping higher education by examining its influence on key areas such as pedagogy, technology acceptance, and the ethical implications surrounding its use. The focus on academic integrity within this theme highlights the urgent need for institutions that create strong rules and plans to guarantee that the educational benefits of AI are harnessed without compromising ethical standards. The study emphasizes that while Generative AI holds transformative potential for enhancing teaching and learning, it must be accompanied by vigilant efforts to safeguard the integrity of academic practices in this new digital era.
Cluster 2 focuses on “The integration of large language models in education and their ethical implications,” which reflects the growing integration of LLM models like GPT-4 in the education sector, focusing on their applications, challenges, and ethical considerations. As LLMs become increasingly embedded in education computing, they are being utilized for a variety of motives, such as assessment, feedback, e-learning, and language education. These models offer innovative solutions for enhancing learning systems and providing personalized educational experiences, particularly for university students.
However, the deployment of LLMs in education also raises significant ethical concerns. Issues like the possibility of biased outputs, the accuracy of automated assessments, and the implications for academic integrity necessitate a critical examination of how these technologies are used. The keywords also highlight the importance of generative AI literacy among both educators and students, ensuring that they understand the capabilities and limitations of these technologies. Moreover, ethical frameworks are desperately needed to direct the responsible use of LLMs in educational contexts as technology advances, balancing their potential benefits with the need to uphold ethical standards and promote fair and effective learning outcomes.
Cluster 3 focuses on “AI-driven educational innovations in medical and nursing education,” which centers on the transformative role of AI in reshaping the landscape of medical and nursing education. Integrating AI technologies, such as NLP language and machine learning, enhances both learning and teaching methodologies in these fields. AI-driven tools are increasingly being used to develop adaptive curricula, improve writing and critical thinking skills, and optimize student performance through personalized learning experiences.
In medical and nursing education, AI supports a more efficient and tailored approach to learning, enabling students to engage with complex materials at their own pace while receiving immediate feedback. These innovations are enhancing traditional educational practices and fostering the development of critical skills essential for healthcare professionals. The use of AI in this context emphasizes the need for a curriculum that incorporates these technologies, preparing students to navigate and excel in a healthcare environment that is increasingly reliant on AI-driven systems. As AI is still developing, its role in medical and nursing education is probable to expand, further enriching the learning experience and better equipping students for their professional careers.
Discussion
This study provides a comprehensive bibliometric analysis of the evolving research landscape surrounding the adoption and impact of generative AI technologies in education, particularly focusing on higher education. By addressing the research questions, we have identified key trends, influential journals and articles, the intellectual relationships between contributing countries, the common keywords used in the field, and the prominent themes and topics that currently define this area of study.
The analysis of annual publication trends reveals a significant surge in research output over the past 10 years, with 2024 showing a remarkable increase in the number of publications compared to other years. This rapid growth reflects the escalating interest and urgency within the academic community to explore the implications of generative AI, especially as these technologies become increasingly integrated into educational practices. The adoption of Gen-AI tools like ChatGPT and GPT-4, particularly in response to the challenges posed by remote learning and digital education, has likely driven this surge, indicating a shift in educational paradigms towards more AI-centered approaches.
The study highlights the most productive journals contributing to this field, with Education and Information Technologies emerging as leading sources of impactful research. Articles that address the implications of ChatGPT on academic integrity, such as those by Rudolph et al. (2023a) and Cotton et al. (2024) have garnered significant attention, underscoring the critical discourse surrounding AI’s role in upholding or challenging traditional educational values. These influential articles serve as foundational texts for ongoing discussions about the ethical and practical aspects of integrating Gen-AI into educational systems.
The bibliographic coupling analysis reveals a strong intellectual network between countries, with the United States, Australia, and the United Kingdom leading in research productivity and influence. These countries demonstrate high citation counts and strong collaborative ties, suggesting that they are at the forefront of exploring and shaping the global discourse on Gen-AI in education. The prominence of countries like China and the United Arab Emirates further highlights the global nature of this research area, with diverse educational systems contributing to the collective awareness regarding the effects of Gen-AI on education.
The keyword analysis reveals a set of commonly used terms such as “ChatGPT,”“Generative AI,”and“artificial intelligence,” which are central to the discourse on AI in education. These keywords are closely associated with themes of educational technology, ethics, and the impact of AI on teaching and learning practices. The analysis also shows that these keywords are distributed across different countries and authors, indicating a wide-ranging interest in exploring different approaches to using Gen-AI tools for better education results while responding to associated ethical challenges.
The co-occurrence analysis of keywords has allowed us to identify several prominent themes in the current research landscape. Themes such as “Generative AI’s Impact on Higher Education and Academic Integrity,”“The integration of large language models in education and their ethical implication,” and “AI-Driven Educational Innovations in Medical and Nursing Education,” reflect the broad application of AI across various educational contexts. These themes underscore the dual focus on technological innovation and ethical considerations, highlighting the potential of Gen-AI to revolutionize education while also raising important questions about its implications for academic integrity and student learning.
Overall, this study has mapped the rapid evolution of research in the field of generative AI in education, providing insights into the key trends, influential contributions, and emerging themes that define this dynamic area of inquiry. As Gen-AI continues to transform educational practices, ongoing research will be crucial in guiding these technologies’ ethical and effective incorporation into learning environments, ensuring that they enhance rather than undermine the educational experience.
Theoretical Implications
This study offers several important theoretical implications that advance our understanding of the generative AI’s role in 21st-century education. First, this study contributes to the technology-enhanced learning and educational technology literature by mapping the trajectory of Gen-AI as a disruptive force shaping pedagogical theories and learning models. The flow in publications and thematic focus on tools like ChatGPT suggests that Gen-AI is a technological tool and a catalyst for rethinking learning theories, including constructivism, connectivism, and personalized learning paradigms. Second, the findings underscore the growing integration of Gen-AI into higher education in explaining how AI tools impact student motivation, autonomy, and information processing. Understand how learners interact with AI-driven content and how educators can adapt teaching strategies accordingly. Third, the study adds a bibliometric layer to the theoretical discourse by identifying the intellectual structure and knowledge clusters within Gen-AI in education. Through co-word and bibliographic coupling analysis, it identifies emerging subfields and interdisciplinary linkages, suggesting the evolution of a new theoretical domain that intersects artificial intelligence, learning sciences, and digital pedagogy.
Fourth, the global distribution of research—led by countries such as the United States, Australia, and the United Kingdom—indicates a geographically uneven diffusion of Gen-AI-related educational innovation, raising theoretical questions about technological equity, digital colonialism, and the contextual adaptability of Gen-AI across diverse educational systems. Fifth, the study highlights the need to provide better interpretation for the difficulties introduced by Gen-AI driven tools like ChatGPT, DALL-E LLaMA2, Gemini, DeepSeek, and Microsoft Copilot. Last, this study also underscores the importance of interdisciplinary approaches, combining insights from educational theory, AI ethics, and cognitive science to develop more holistic models that can guide the effective and responsible incorporation of Gen-AI in education.
Practical Implications
This study has important implications for educators, legislators, and institutions of higher education as they adapt to the growing presence of generative AI in the academic landscape. The flow in Gen-AI-related publications indicates growing acceptance and usage in education. Institutions should incorporate Gen-AI tools like ChatGPT into curricula to enhance personalized learning, critical thinking, and digital literacy. For educators, the study underscores the importance of incorporating AI tools like ChatGPT into curriculum design to enhance student engagement and personalize learning experiences. Educators can foster critical thinking and improve academic performance by incorporating AI-driven technologies into teaching practices. Policymakers and institutions are urged to revise academic integrity protocols in response to the challenges posed by gen-AI, such as increased risks of plagiarism and cheating. This necessitates implementing robust AI detection tools and clear guidelines on ethical AI use. Additionally, the study highlights the need for ongoing professional development for educators, ensuring they are equipped with the skills to effectively integrate AI into their teaching and guide students in its responsible use.
Teachers and academic staff need ongoing professional development to integrate Gen-AI tools into pedagogy effectively. Regional disparities in adoption, particularly between Western countries (e.g., United States, Australia, and United Kingdom) and regions like the UAE or China, highlight the importance of context-specific training and policy design. Governments and academic institutions must create ethical frameworks and data governance policies to manage AI usage in classrooms responsibly. Comparative insights suggest that Western countries are advancing faster due to supportive regulatory and funding environments, offering models for other nations. Furthermore, the study calls for the development of comprehensive policies and ethical standards to address issues like data privacy and algorithmic bias in AI applications. Finally, promoting AI literacy among students is essential, equipping them with the knowledge to use AI tools effectively and ethically. These practical implications provide a roadmap for stakeholders to harness the benefits of generative AI while mitigating its challenges in the education sector.
Limitations and Future Research Directions
While providing valuable insights into the generative AI and trends, challenges, and benefits in education, although this study has several limitations. First, the limitation is the reliance on bibliometric data from a specific timeframe (2015-2025), which may not fully capture the evolving nature of AI technologies and their applications in education. For future research, it is recommended to extend the timeframe of analysis and include a broader range of languages and regions to gain a more holistic view of global trends. Second, this study focuses primarily on articles written in English, potentially overlooking significant contributions from non-English-speaking regions, which could provide a more comprehensive understanding of global trends. Future research would consider both English and non-English regions. Third, this study employs a bibliometric analysis using data from WoS and Scopus. Future longitudinal studies could provide deeper insights into the long-term effects of AI integration in education. Fourth, a rapidly changing technological landscape, this study mainly focused on Gen-AI based technologies and academic performance-based keywords. Future studies would consider different AI technologies to explore findings. Fifth, the use of co-occurrence and bibliographic coupling analyses, while effective for identifying relationships and trends, may not delve deeply into the qualitative aspects of how Gen-AI is transforming educational practices. Furthermore, qualitative research approaches, such as case studies or interviews with educators and students, could offer more understanding of these technologies. Last, exploring ethical considerations, this study included keywords ethical challenges and the implications for academic integrity. Future bibliometric analyses could use bias, misinformation, and data privacy-related keywords for deeper exploration.
Footnotes
Acknowledgements
The authors thank the editors and anonymous reviewers for their valuable feedback to improve the quality of this work.
Ethical Considerations
Not applicable, as this study did not involve any human/animal data or pictures.
Consent to Participate
Not applicable, as this study did not involve any human/animal data or pictures.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by National Natural Science Foundation of China [Grant numbers 72474016, 72004012, and 72074014].
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
Data will be made available on request.