A Systemic Functional Linguistic Analysis of ChatGPT’s Academic Writing Responses: Metafunctional Patterns and Implications for Writing Pedagogy

Abstract

Generative artificial intelligence (AI), particularly ChatGPT, is having a significant impact on how academic writing is produced, taught and assessed in higher education. This study uses Systemic Functional Linguistics (SFL) to analyze how ChatGPT constructs meaning when it responds to academic writing prompts. Drawing on Halliday’s three metafunctions (ideational, interpersonal and textual), it examines ten ChatGPT-generated texts representing five academic part-genres: the research introduction, the argumentative essay, the literature review, the methodology section and the discussion section. The analysis shows that ChatGPT produces highly formal, register-appropriate academic discourse characterized by dense nominalization, median- and low-value modality and formulaic evaluation. It makes sophisticated use of transitivity, mood and thematic progression, yet it constructs an “epistemic space of safe neutrality” in which alternative views are acknowledged but, where argument is most expected, left unresolved. Cross-genre analysis indicates that discussion sections have the highest modal density and nominalization and the most heavily hedged interpretive claims, argumentative essays leave their concessions unresolved, and methodology sections show the most procedural sequencing. These findings have implications for writing pedagogy and assessment at a time when machines can reproduce the form of scholarship with great fidelity. The study argues that SFL gives teachers an explicit metalanguage for educating students as critical consumers of language, able not only to produce texts but also to understand how meaning is constructed, negotiated and contested in and through language.