Stance and Authorial Voice in Human and AI-Generated Academic Texts: A Corpus-Based Discourse Analysis

Stance and authorial voice constitute essential features of academic discourse, allowing writers to present claims, express evaluation, and position themselves within the scholarly conversation. Academic abstracts are central to scholarly communication, as they not only summarize research but also construct authorial stance and voice. With the growing use of artificial intelligence in academic writing, it is important to examine how AI-generated abstracts conform to established discourse conventions. This study investigates stance and authorial voice in human-written and AI-generated academic abstracts through a corpus-based discourse analysis. The current study uses Hyland’s interactional model of stance. A small, specialized corpus of linguistics-related abstracts that were human-written and AI-generated was collected. That was manually analyzed to examine the use of hedges, boosters, attitude markers, and self-mention. The findings indicate that while both human and AI-generated abstracts employ core stance resources typical of academic discourse, notable differences exist in their discursive deployment. Human-written abstracts demonstrate greater rhetorical control and contextual sensitivity, whereas AI-generated abstracts rely on more uniform and formulaic stance patterns. These differences are primarily discursive rather than grammatical, highlighting the role of rhetorical intentionality in academic abstract writing and contributing to discussions on AI-mediated scholarly communication.