LINGUOCULTURAL AND METADISCURSIVE TRANSFORMATION IN AI-ASSISTED ACADEMIC WRITING: A CONTRASTIVE STUDY OF ENGLISH AND UZBEK EFL WRITERS

Abstract

This article investigates how generative artificial intelligence may transform linguocultural meaning and metadiscursive choices in academic writing by English-language learners with Uzbek linguistic backgrounds. Drawing on contrastive rhetoric, intercultural pragmatics, and interactional models of metadiscourse, the study develops an analytical framework for comparing independently written English texts, AI-revised English texts, and culturally situated Uzbek academic-writing conventions. A qualitative, literature-informed contrastive design is adopted; illustrative textual pairs are used to demonstrate coding procedures rather than to represent empirical participant outcomes. Five dimensions are examined: stance and hedging, reader engagement, authorial visibility, cultural framing, and rhetorical organization. The analysis suggests that AI-assisted revision can improve surface coherence and conventional academic phrasing while potentially neutralizing culturally significant expressions, shifting epistemic commitment, and homogenizing writer voice. A human-in-the-loop protocol is proposed to help EFL writers evaluate AI suggestions without losing rhetorical agency. The article contributes a replicable framework for future English-Uzbek corpus studies and pedagogical research.