Generative-AI/LLM-Enabled E-Assessment in Higher Education: A Bibliometric Analysis of Research Trends, Collaboration Networks, and Thematic Evolution

This study conducts a comprehensive bibliometric analysis of research on generative-AI/LLM-enabled e-assessment in higher education, drawing on 243 peer-reviewed journal articles published between 2023 and 2026 and indexed in the Scopus database. Employing VOSviewer and Bibliometrix software tools, the analysis integrates performance metrics and science mapping techniques. The findings reveal rapid growth, with annual publications increasing from 4 in 2023 to 117 by July 2026 (H-index = 35; 3854 total citations). China leads scientific production (39 articles), while Assessment and Evaluation in Higher Education is the most prolific journal (14 articles). The most cited study addresses LLM-generated feedback effectiveness (494 citations). VOSviewer keyword co-occurrence analysis identifies four thematic clusters: (1) generative AI and feedback processes, (2) health-professions education and methodological designs, (3) AI in education and pedagogical systems, and (4) large language models and automated assessment. The co-authorship network reveals a dispersed collaboration structure with 28.0% international co-authorship. Thematic evolution analysis reveals a shift from foundational explorations of AI viability toward specialized applications in feedback literacy and self-regulated learning. Several research gaps are tentatively identified as hypotheses for future verification, including the underrepresentation of developing countries, limited attention to ethical dimensions, and a scarcity of longitudinal studies.