In this systematic review, we review the transformations caused by artificial intelligence in higher education in terms of educational quality and digital equity. A systematic search was conducted in five multidisciplinary databases—Scopus, Web of Science, ScienceDirect, ERIC, and Taylor & Francis Online—using PRISMA 2020, covering studies from 2018 to 2025. After applying strict inclusion and exclusion requirements, we selected 50 studies with a minimum level of methodological quality. Our results show that AI functions in higher education in dialectical tension, as it operates as both a driver of academic personalization and an amplifier of structural inequalities. Our main results identify five applications: intelligent tutoring systems, adaptive assessment platforms, predictive learning analytics, academic recommendation systems, and administrative automation. Intelligent tutoring systems show improvements in student retention, and adaptive assessment systems show advances in personalised assessment and feedback. However, our analysis also uncovers biassed and systematic algorithmic practices that negatively affect culturally diverse populations and students from the global south. The high concentration of planned research in the global north is an epistemological bias and a limitation to the cross-cultural transferability of results. Research contributes to the development of an integrated educational AI framework for higher education that addresses the pedagogical, technological, ethical, and distributional dimensions. A key finding is that, in the absence of a careful examination of structural determinants and a comprehensive ethical framework, and democratic participation in design processes to create educational equality, technology alone is insufficient.