The Post-ChatGPT Research Landscape of Generative Artificial Intelligence in Higher Education (2023–2026): A Bibliometric Analysis - Figure 1

To clearly illustrate the study selection process, the steps described above were synthesised in a PRISMA-type flow diagram (see Fig. 1), in accordance with the PRISMA 2020 recommendations for transparent reporting of the record inclusion process and their adaptation to bibliometric analyses (Page et al., 2021). In the first stage, a total of 36,483 records were identified in the Web of Science database using the predefined search strategy, which combined a Topic (TS) query with a temporal limitationto the period 2023–1 February 2026, the document types “Article”, “Review”, and “Early Access”, and a restriction to publications in English. In the second stage, the 36,483 records underwent a systematic screening process based on document type and relevance criteria aligned with the study’s objective. During this stage, 15,828 records were excluded, mainly books, book chapters, conference proceedings papers, and retracted publications that did not correspond to the proposed bibliometric design After applying the inclusion and exclusion criteria, 20,655 records met the eligibility criteria for inclusion in the third stage and were retained for further analysis. The eligible corpus consisted exclusively of documents classified as “Article”, “Review Article”, and “Early Access”, to ensure the comparability and quality of the analysed data. In the final stage, the 20,655 eligible studies were included in the bibliometric analysis itself. The corpus comprised 16,961 original research articles (Articles), 1,921 publications classified as “Early Access”, and 1,773 review articles (Review Articles), providing a comprehensive overview of recent scientific output in the investigated field.