A Corpus-Based Comparative Analysis of Human and ChatGPT-Generated Academic Abstracts

Abstract

Academic writing continues to evolve alongside advances in digital technology, particularly the growing use of artificial intelligence in research communication. However, limited linguistic research has compared human and AI-generated academic texts to identify measurable differences in writing quality. This study analyses a corpus of 20 academic abstracts to compare lexical and syntactic features across both authorship types. Using qualitative and corpus-based methods, the study examines vocabulary range, academic terminology, sentence structure, and clause complexity. The findings of the present study reveals that human abstracts demonstrate greater lexical diversity, disciplinary precision, and syntactic depth, while ChatGPT abstracts display clearer structure but increased repetition and simplified syntax. The study concludes that although AI generates fluent academic language, it does not yet replicate the linguistic nuance found in human academic writing.