AI-Powered Learning Content Generation through Retrieval-Augmented Generation for Improved Accuracy and Personalisation

Abstract

The integration of artificial intelligence (AI) in education has transformed content delivery and personalisation, with
Retrieval-Augmented Generation (RAG) emerging as a promising architecture to enhance the accuracy and contextual
relevance of AI-generated educational materials. This study aims to systematically review the application, effectiveness,
and challenges of RAG-powered educational content generation systems. The research employed a systematic literature
review (SLR) design, guided by the PRISMA 2020 protocols. A total of 26 peer-reviewed articles published between
2020 and 2025 were selected from the Scopus database. No human participants were involved, as the study is based on
literature. Data were extracted using a predefined matrix and analysed thematically to identify recurring patterns and
contradictions. Findings revealed three major themes: (1) the dominance of retriever-generator pipelines and modular
platforms such as OpenRAG and LearnRAG; (2) significant improvements in content factuality, student performance,
and explainability; and (3) persistent limitations including infrastructure constraints, data bias, and ethical concerns.
The review concludes that while RAG enhances educational AI systems, equitable access and responsible implementation
remain critical. Implications include the need for policy frameworks, improved infrastructure, and future research on
multilingual and domain-specific RAG applications.