Rethinking Educational Scaffolding in Language Learning: A Theory-Driven Critical Review of Large Language Models

Generative Artificial Intelligence (GenAI), particularly large language models (LLMs), has rapidly emerged as a transformative technology in language education by providing learners with personalized feedback, interactive dialogue, adaptive explanations, and continuous learning support. An increasing body of research has explored the application of LLMs across different language learning contexts and skills; however, the role of LLMs as educational scaffolds remains conceptually and theoretically underexplored. Although many studies describe LLM-based interventions as forms of scaffolding, there is limited understanding of how educational scaffolding is conceptualized and operationalized in existing research, and whether the scaffolding functions of LLMs align with established theories of educational scaffolding. Furthermore, current literature demonstrates considerable variation in pedagogical approaches, methodological designs, and theoretical foundations, creating a need for a comprehensive synthesis of existing evidence. This project is a systematic critical review that aims to identify how LLMs have been utilized to support educational scaffolding in language learning, examine how scaffolding is defined and implemented, and critically evaluate the alignment between LLM-mediated support and established scaffolding theories. The review will follow a predefined protocol involving a comprehensive search strategy, explicit eligibility criteria, systematic study selection, and transparent data extraction procedures. Reporting will adhere to the PRISMA 2020 guidelines. Beyond summarizing existing findings, this review will critically synthesize the pedagogical strengths, limitations, theoretical assumptions, and methodological gaps within the current literature. The review will ultimately develop a theory-informed conceptual framework that explains the role of large language models in educational scaffolding and provides directions for future research, educational practice, and the responsible integration of AI technologies into language learning environments.