The scale of the analysis required a rigorous screening pipeline. The researchers began with 2,359 records retrieved from Scopus on December 27, 2025, using a query that matched the terms artificial intelligence or AI against inclusive education in titles, keywords and abstracts. Filtering to the 2021–2025 window left 2,286 documents, restricting the language to English reduced the pool to 2,246, and manual screening for topical relevance ultimately retained 1,655 documents. The final corpus spanned six document types: 542 journal articles, 454 conference papers, 449 book chapters, 105 reviews, 83 books and 22 conference reviews. Scopus was chosen deliberately, the authors note, because it offers roughly 20 percent greater citation coverage than Web of Science, making it well suited to mapping a young, fast-moving, multidisciplinary field.

Two complementary software tools powered the analysis. Bibliometrix, an R package built for quantitative science mapping, was used to identify the most relevant authors, institutional affiliations, corresponding-author countries and most-cited nations. VOSviewer, a free network-visualization program, was deployed to chart bibliographic coupling among sources, countries and authors, revealing the hidden intellectual architecture of the field. Together, the tools allowed the team to move beyond simple publication counts and reconstruct how ideas, citations and collaborations flow through the research landscape of AI for inclusive education.

Citation analysis identified the field’s intellectual anchors. The most highly cited work, by Chen and colleagues on AI student assistants and chatbot design for classroom success, accumulated 431 citations, while Southworth and colleagues’ framework for embedding AI literacy across higher education curricula followed with 389. The most-cited studies clustered around a recognizable set of themes: AI literacy, adaptive learning systems, generative artificial intelligence, chatbot-assisted learning, personalized learning environments, ethical governance and institutional policy frameworks. Since 2023, the authors found, academic curiosity has pivoted sharply toward generative AI, with ChatGPT and large language models dominating the high-citation literature and raising parallel questions about assessment, curriculum design and responsible use.

The subject-area breakdown underscores how fundamentally interdisciplinary this field has become. Computer Science leads with approximately 26.6 percent of publications, followed closely by Social Science at 25.6 percent, a pairing the authors read as evidence that technological development and pedagogical, equity-oriented perspectives are advancing in tandem. Engineering contributes 9.9 percent, while Business, Management and Accounting (4.9 percent), Mathematics (4.8 percent) and Medicine (4.2 percent) add moderate but meaningful weight. Smaller shares come from Arts and Humanities, Economics, Decision Sciences and Psychology, with other fields accounting for 12.2 percent. No single discipline owns the topic, and the authors argue that successful real-world implementation will demand coordinated effort across all of them.

Geographically, the field is strikingly concentrated. The United States, India and China emerged as the leading producers, with India and the United States each contributing around 580 and 566 documents respectively, followed by Indonesia with 204 and China with 201. The United States also dominates citation impact with 1,892 citations, ahead of China with 667 and the United Kingdom with 602. At the institutional level, Inti International University tops the list with 16 publications, followed by King Faisal University with 11, while Qatar University, the University of California and the University of Florida contributed nine each. On authorship, Asrifan A leads with 12 articles, most of them published in 2025, with Bozkurt A, Chen Y, Li J and Gupta S tied at eight publications each.

Bibliographic coupling networks added a structural dimension to these rankings. Source-level analysis revealed five clusters, with Lecture Notes in Computer Science showing the highest link strength at 119, followed by Frontiers in Education at 73 and the International Journal of Learning, Teaching and Educational Research at 69. Country-level coupling produced six clusters, with the United States again leading at 4,897 link strengths across 300 documents, followed by India at 4,644 and the United Kingdom at 2,746. Author-level networks formed three main clusters, anchored respectively by Asrifan, Acosta-Vargas and Compitiello; Dimitriadis and Laverde; and Roe, Al Shahrani and Jemni. Asrifan recorded the highest individual link strength at 198, with Bozkurt close behind at 177 alongside 102 citations.

Beneath the growth curves, the study exposes significant gaps. The most-cited research disproportionately targets higher education, universities, undergraduate learners and institutional policy, while disability-centered inclusive education, school-level settings and primary learners receive comparatively little attention. Geographically, technologically advanced nations dominate the citation landscape while developing regions remain underexplored. Long-term instructional analyses are scarce, with most studies examining only short-term learner engagement, and socio-cultural impacts of AI integration remain inadequately investigated. The authors also observe a shift in the field’s center of gravity: from performance-focused adaptation toward equity-centered adaptation, echoing the United Nations Sustainable Development Goal 4 on quality education, though they caution that inclusion-related themes are still unevenly integrated across research clusters.

The study’s limitations deserve equal attention. It relies solely on Scopus, excluding relevant work indexed in Web of Science, ERIC, Google Scholar and regional databases. Only English-language publications were included, omitting meaningful contributions from non-English-speaking contexts. The 2021–2025 window cannot capture foundational earlier work, and bibliometric methods inherently privilege quantity and citation popularity over qualitative depth, meaning highly cited studies are not always the highest quality. Narrow search terms built around artificial intelligence, AI and inclusive education may also have missed related literature using alternative vocabulary, and rapidly evolving recent developments may not yet appear in citation databases.

For educators, researchers and policymakers, the message is clear. Generative AI tools, chatbots, intelligent tutoring systems and adaptive learning platforms hold genuine potential to personalize instruction, boost engagement and widen access for learners with diverse needs and abilities. But realizing that potential, the authors conclude, depends on pedagogical design, teacher involvement and theoretical grounding in equity and participation rather than technology alone. They call for system-wide AI literacy frameworks, adaptive governance structures addressing ethics, assessment integrity and data protection, and sustained investment in digital infrastructure so that AI-assisted pedagogies complement rather than replace human teaching. As publication output continues its steep climb, the field’s challenge is no longer generating interest in AI for inclusive education, but ensuring that interest translates into fair, ethical and globally balanced practice.

Subject of Research: Bibliometric analysis of global research trends in artificial intelligence for inclusive education from 2021 to 2025

Article Title: A bibliometric study of publication trends and intellectual structure in artificial intelligence for inclusive education

Article References: Baro, K., Das, B., Basumatary, S., & Mochahary, D. (2026). A bibliometric study of publication trends and intellectual structure in artificial intelligence for inclusive education. Discover Education, 5(1), Article 995. https://doi.org/10.1007/s44217-026-01794-2

Image Credits: AI Generated

DOI: 10.1007/s44217-026-01794-2

Keywords: artificial intelligence, inclusive education, bibliometrics, generative AI, adaptive learning, chatbots, personalized learning, AI literacy, Scopus, VOSviewer, educational technology, research trends