Bibliometric mapping of generative artificial intelligence research in higher education using VOSviewer and CiteSpace
TL;DR
The findings suggest that future research and practice should focus on how generative AI can be used effectively, responsibly, and sustainably in authentic higher education settings, with attention to learning quality, long-term effects, fairness, data ethics, and governance.
Abstract
This study aimed to map research trends, knowledge structures, and emerging frontiers of generative artificial intelligence in higher education. Based on 774 publications retrieved from the Web of Science database from 2021 to 2026 (Retrieval Date: May 16, 2026), VOSviewer and CiteSpace were used to analyze publication trends, core authors, journals, institutions, keyword co-occurrence, thematic clusters, co-citation networks, keyword time zones, and burst terms. The results show rapid growth after 2023, with a publication peak in 2025. Keyword co-occurrence and clustering analyses identify four interconnected themes: educational processes and learner behavior, technology acceptance and use, AI technology applications and methods, and educational outcomes and assessment. Co-citation and temporal analyses indicate that the field is supported by research on ChatGPT and large language models, technology acceptance theories, and application-oriented studies in contexts such as language learning, academic writing, medical education, assessment, and feedback. The research frontier is shifting from tool adoption and student perceptions toward cognitive load, self-regulated learning, learning analytics, feedback mechanisms, assessment redesign, academic integrity, and institutional policy. These findings suggest that future research and practice should focus on how generative AI can be used effectively, responsibly, and sustainably in authentic higher education settings, with attention to learning quality, long-term effects, fairness, data ethics, and governance.