Syntactic complexity of student-produced, AIedited, and AI-generated expository texts using ChatGPT-4.0: evidence from Vietnamese and Filipino EFL undergraduates

Abstract

While studies have explored syntactic complexity in second-language (L2) writing, little is known about the syntactic features of AI-generated and AI-edited texts, or how these texts compare with student-produced essays. Thus, this study sought to fill in this gap by analyzing 690 expository essays (i.e. 230 essays produced by Vietnamese and Filipino students, 230 AI-edited essays, and 230 AI-generated essays) using 14 syntactic complexity measures. Our study revealed significant syntactic differences across the three groups in all 14 syntactic complexity measures, with student-produced essays generally exhibiting higher syntactic complexity, while AI-generated essays showed increased syntactic density but simpler overall structures. AI-edited essays fell between the two, enhancing certain syntactic aspects without fully replicating the complexity of human-written texts. Implications for L2 writing pedagogy, AI tool enhancement, and future studies are discussed.