Generative Artificial Intelligence in Education From 2021 to 2025: A Scientometric Review

Abstract

This study presents a bibliometric analysis of 965 peer-reviewed articles on generative artificial intelligence (GenAI) in education published from 2021 to 2025 in the Web of Science Core Collection. Through keyword co-occurrence, co-citation, and collaboration network analyses, it identifies core research themes, intellectual structures, and developmental trends. Findings reveal an exponential rise in GenAI-related publications, with dominant themes centred on technological applications of GenAI in teaching and assessment—especially ChatGPT—alongside technology acceptance mechanisms and learner outcomes such as motivation and self-efficacy. Three major thematic clusters emerge: GenAI educational applications, user adoption theories, and learning impacts. Co-citation patterns show strong reliance on traditional acceptance models like TAM, indicating limited development of GenAI-specific theoretical frameworks. Collaboration analyses reveal fragmented author networks and uneven global participation, concentrated mainly in North America and East Asia. The study highlights research gaps, including ethical governance, creativity development, interdisciplinary applications, and insufficient qualitative or mixed-method studies. It recommends developing theoretical models tailored to GenAI’s interactive and multimodal characteristics, strengthening ethical and cross-cultural frameworks, expanding interdisciplinary innovation, and enhancing global research cooperation to support the sustainable and responsible integration of GenAI in education.