Using dynamic feedback from customized LLM-powered chatbots in EFL argumentative writing

With the growing use of large language models (LLMs) in English as a foreign language (EFL) teaching, balancing teacher control over instructional focus with students’ autonomy in using artificial intelligence chatbots while ensuring more dynamic interaction have become key pedagogical concerns. This article reports on an innovative teaching practice implemented in a foundation undergraduate English for Academic Purposes course, which featured interactive feedback, including chatbot-mediated dynamic assessment, delivered through two custom-built LLM-powered chatbots. The intervention aimed to support students’ argumentative writing with specific reference to argumentation quality and linguistic accuracy, through an iterative multi-draft process. Drawing on observation and critical reflection on this practicum, the article argues that customized LLM-powered chatbots can provide dynamic, instruction-aligned feedback at scale to scaffold EFL learners’ writing process, while instructors remain essential in the loop to address the limitations of automated feedback. The article concludes with suggestions for refining this approach in future practice.