A Systematic Review of Large Language Models in Education Dual Roles, User Perspectives, and Cross-Disciplinary Impacts

LLMs are increasingly shaping the landscape of education. However, comparative analyses of their functional roles and educational impact across different perspectives remain limited. This systematic review examines 82 peer-reviewed articles (2023–2025) to explore the dual roles of LLMs—as learners and instructors—and assess their educational impact from student and educator perspectives. As learners, models such as ChatGPT-4 and ChatGPT-4o demonstrate high accuracy across disciplines, including medicine, programming, and others, particularly in standardized assessments. As instructors, LLMs show strong potential in automatic grading and related functions.Additionally, their broader educational impact is mixed. Learners benefit from personalized feedback and study assistance, while educators leverage LLMs for curriculum planning and content creation. Nonetheless, concerns persist regarding overreliance, misinformation, and academic integrity. This review highlights the current capabilities and limitations of LLMs, offering insights to guide future research and support their responsible adoption in education.