Systemic Functional Perspectives on Theme and Rheme in AI-Assisted EFL Writing

Abstract

Artificial intelligence-assisted writing tools are increasingly used in EFL classrooms to support idea generation and text production. However, their contribution to discourse-level organization remains underexplored, particularly in relation to Theme–Rheme structures and thematic progression from a Systemic Functional Linguistics perspective.This study employed a qualitative descriptive design to examine Theme–Rheme organization in AI-assisted EFL narrative writing. The data consisted of 31 narrative texts produced by undergraduate English Education students using the Storybird platform. The texts were analyzed clause by clause using Halliday’s Systemic Functional Linguistics framework. Each clause was coded to identify topical, textual, and interpersonal Themes, as well as their role in shaping discourse organization.The analysis identified 520 Themes across the student texts. Topical Themes were dominant, accounting for 438 occurrences, followed by 73 textual Themes and 9 interpersonal Themes. The findings indicate that students mainly relied on unmarked topical Themes and subject–predicate structures to organize their narratives. Although this pattern supported clause-level clarity, the limited use of textual and interpersonal Themes suggests restricted thematic variation and limited control over discourse cohesion, stance, and rhetorical development.The findings show that AI-assisted writing can support idea generation but does not automatically develop students’ discourse-level competence. Students still require explicit instruction in Theme–Rheme organization and thematic progression to produce coherent and rhetorically effective texts. Integrating AI-based writing tools with SFL-informed pedagogy may help EFL learners move beyond grammatical accuracy toward stronger discourse organization.