Rhetorical Effects of AI Paraphrasing on Argumentation and Authorial Stance in Arab EFL Literature Reviews

Abstract

Recent research has identified systematic differences between AI-generated and human-authored academic writing. However, most studies focus on argumentative essays, leaving the rhetorical effects of AI use on literature reviews underexplored. This paper addresses this gap through a controlled analysis of how AI paraphrasing reshapes argumentation and authorial stance in Arab EFL postgraduate literature reviews under content-preserving conditions.

Using a paired corpus design, this study compares ten human-authored literature reviews produced in 2020, prior to the widespread use of large language models, with ChatGPT-paraphrased versions produced using a controlled prompt. This design isolates core AI transformation patterns. Analysis draws on Toulmin’s model of argumentation and Hyland’s stance framework, supported by corpus-based validation.

Results indicated systematic rhetorical transformation. While ChatGPT preserved structural coherence and improved surface fluency, the AI-generated texts showed reduced rhetorical complexity and weakened authorial stance. Self-mention and reader inclusion were sharply reduced, hedging became more formulaic, warrants shifted from situated reasoning to generic principles, and contrastive and causal connectives occurred less frequently. These findings clarify which rhetorical features are most vulnerable under controlled AI paraphrasing and position AI paraphrasing as a concern for EFL academic writing pedagogy.