The impact of RAG chatbots on writing proficiency: A study of Vietnamese Grade 6 EFL students

This study investigates the use of Retrieval-Augmented Generation (RAG) chatbots in writing instruction for Grade 6 EFL students in Vietnam, with a focus on their impact on students’ writing proficiency. Employing a mixed-methods quasi-experimental design, the research involved 70 students from the Asian International School (AIS) in Ho Chi Minh City, who were divided into two groups: a control group (n = 34) that received traditional instruction and an experimental group (n = 36) that engaged with RAG chatbots during writing activities over a 10-week period. Writing proficiency was assessed through pre- and post-tests scored with an IELTS-adapted rubric. Quantitative findings showed that the experimental group achieved significantly greater writing gains than the control group (M gain = 0.39 vs. 0.09, p < .05; d = 0.59), suggesting that RAG chatbot-supported instruction was associated with improved writing proficiency. Qualitative analysis of interviews and chatbot logs revealed that students benefited from lexical and grammatical enrichment, procedural scaffolding, iterative engagement, and affective support which further supported the positive impact of the RAG chatbot. This study is one of the few to explore RAG chatbot integration in Vietnamese secondary education. By implementing a curriculum-aligned RAG chatbot in an authentic LMS-based AIS Grade 6 EFL classroom, the study demonstrates the potential and significance of RAG technology as a scalable pedagogical support for developing writing proficiency in EFL and similar technology-supported school settings.