AI-Generated Content in Education: Moving Beyond Theory toward Data-Driven and Actionable Insights

Abstract

The emergence of AI-powered content such as text, images, and videos generated by large language models (LLMs) has rapidly transitioned from an experimental novelty to a structural force in higher education. Yet, previous reviews have remained largely conceptual or focused on the early stages of the AI wave. This study offers an updated, multi-faceted overview of AI-generated content in education based on data from 2022 to 2025. Following the PRISMA protocol, we systematically identified and reviewed 64 peer-reviewed studies to examine usage, transformation of teaching practices, impacts on learning outcomes, and challenges related to academic integrity, bias, trustworthiness, and teacher competence. Our findings reveal a clear progression: AI content is no longer on the sidelines but is reshaping assessment design, pedagogical strategies, and learner agency. However, this involves significant risks, including tool dependency, inequality, and a lack of regulatory framework. The study proposes specific recommendations for integrating AI responsibly and humanely into contemporary education.