A bibliometric mapping of research on the ethics of artificial intelligence in education, 2015 to 2026

Research on the ethics of Artificial Intelligence (AI) in education has expanded rapidly, yet its intellectual structure and the effect of generative AI on its priorities have not been mapped at scale. This study analyzed 5,988 documents published between 2015 and October 2026, retrieved from the Semantic Scholar Academic Graph, enriched with Crossref metadata, and validated for precision (87.5%). Performance analysis, co-citation clustering of 181 highly cited references, co-word strategic diagrams for three periods (2015 to 2022, 2023 to 2024, and 2025 to 2026), and length-adjusted logistic models of principle salience were combined. Output grew at a compound annual rate of 111% between 2019 and 2025, and 77.4% of the corpus appeared after 2024. The intellectual base comprised five clusters, ranging from normative and sociotechnical foundations to the institutional governance of generative AI. Academic integrity moved from a small theme to the field's dominant motor theme, learning analytics privacy was absorbed into a governance theme, and algorithmic fairness narrowed into a cohesive but only moderately central theme on trustworthy analytics. Compared with 2015 to 2022, documents from 2025 to 2026 had higher odds of addressing over-reliance (adjusted OR = 10.66), academic integrity (4.79), sustainability (4.00), human agency (2.87), and accountability (2.58), and lower odds of addressing surveillance (0.43). Equity never consolidated into a cohesive theme, although the share of documents with Asian and African authors increased. The field needs to reconnect integrity governance with data rights, distributive justice, and evidence on learning.