A Corpus-Based Comparative Genre Analysis of Human-Written and AI-Generated Research Abstracts: Examining Rhetorical Move Structures

Abstract

Research abstracts are disciplinary and communicative genres that are structured in the form of academic texts. As the academic writing based on the application of Artificial Intelligence tools becomes more common, the questions about the degree to which the AI-generated abstracts can be described by the traditional rhetorical approaches emerge. The purpose of the study is to compare the rhetorical move structure of human-written and AI-generated research abstracts based on the corpus-based comparative genre analysis. The research was based on Swales Move Analysis framework and formed two balanced corpora (40 abstracts by a human and 40 by an AI) under controlled conditions wherein both produced abstracts were generated with the help of the same titles. It used a mixed-method approach that involved qualitative identification of moves with quantitative distribution of frequency and sequencing. Data were gathered using purposive sampling of indexed journals in the same field and during the same period, as well as by using standardized prompts to generate abstracts on AI to be consistent. The results show that both corpora are organized according to the five-move structure, but AI-generated abstracts are less specific in their methods and techniques, have less detailed results descriptions, and are more predictable in linear sequence formation. Although AI systems are effective at reproducing structural conventions, human abstracts are more rhetorically flexible as well as empirically rich, which means that they are more competent in the genre. The paper has significant implications about the academic writing pedagogical method and the use of AI in academic writing.