Artificial Intelligence in Education: A Systematic Review and Bibliometric Topic Modeling Analysis of Applications, Challenges, and Future Directions

Artificial intelligence (AI) has moved from a peripheral experimental technology to a central feature of contemporary educational discourse, prompting a rapidly expanding but fragmented body of scholarship. This paper presents a systematic review combined with a bibliometric and topic-modeling analysis of AI-in-education (AIEd) research published between 2010 and 2021. Following PRISMA-guided screening of 3,938 records identified across Scopus, Web of Science, ERIC, and IEEE Xplore, 158 studies met inclusion criteria and formed the analytic corpus. Bibliometric mapping (co-authorship, keyword co-occurrence, and citation analysis) was performed alongside Latent Dirichlet Allocation (LDA) topic modeling of abstracts to surface latent thematic structures. Results show an accelerating publication trajectory, with output roughly doubling every two to three years after 2016, concentrated in a small set of journals (e.g., Computers & Education, British Journal of Educational Technology) and countries (United States, China, United Kingdom, Australia). Seven latent topics emerged, spanning intelligent tutoring and adaptive learning, learning analytics and educational data mining, conversational agents, assessment automation, personalized learning pathways, AI policy and ethics in higher education, and evolving teacher roles. The most frequently reported applications were intelligent tutoring systems, adaptive learning platforms, and learning analytics/predictive dashboards, while the most cited challenges concerned data privacy and ethics, teacher readiness, and equity in access. The review synthesizes these findings into an integrated conceptual map and outlines a future research agenda emphasizing
explainability, teacher-AI collaboration, longitudinal effectiveness evidence, and equity-oriented implementation research. .