Syntactic and rhetorical transformation in Vietnamese EFL students’ academic writing

Abstract

This study examines how AI-assisted writing tools reshape the linguistic and rhetorical profile of Vietnamese EFL students’ academic texts. Using a convergent mixed-methods design, we analyzed paired drafts (pre-AI and post-AI revision) from 51 English-major undergraduates at two institutions. Quantitatively, syntactic development was gauged with standard indices mean length of T-units (MLT), clauses per sentence (C/S), complex noun phrases per clause (CN/C), and dependent clauses per clause (DC/C) while cohesion was assessed through frequency and range of cohesive ties. Qualitatively, discourse analysis focused on stance, hedging, self-mention, and authorial positioning, complemented by short reflection logs documenting students’ acceptance, adaptation, or rejection of AI suggestions. Results showed statistically significant gains across all syntactic indices and broader, more consistent deployment of cohesive devices. Rhetorically, revisions trended toward greater formality and conventionalized academic tone, often accompanied by increased hedging; however, some dilution of personal voice and culturally situated expression was observed. Reflections revealed diverse interaction patterns with AI, from wholesale acceptance to selective and critical uptake. We argue that AI can function as a productive scaffold for academic writing when embedded in pedagogy that explicitly cultivates critical AI literacy and preserves learner agency. Implications for curriculum design and assessment in Vietnamese higher education are discussed.