A CORPUS-BASED COMPARATIVE ANALYSIS OF SYNTACTIC COMPLEXITY AND STRUCTURAL PATTERNS IN HUMAN-WRITTEN AND AI-GENERATED ACADEMIC TEXTS

Abstract

The swift growth of artificial intelligence has already impacted the academic writing sphere greatly, casting doubt on the language quality and originality of the texts created by the AI. The proposed research aims to perform a comparative syntactic complexity and structural pattern analysis between human-written and AI-generated academic texts using corpus analysis. Based on positivist research paradigm and corpus linguistics and syntactic complexity paradigms, the research study is a mixed-method study that combines both quantitative and qualitative analyses. Purposive sampling was used to create a balanced set of 80 academic texts (40 human written and 40 generated by AI). The sources of data were peer-reviewed academic sources and artificial intelligence-generated productions via standard prompts. Syntactic measures as the mean length of sentence, the density of clauses, and the proportion of subordinates, as well as qualitative assessment of structural arrangements and discourse organization, were used in the analysis. The results indicate that writings by humans have much more advanced syntactic complexity, more embedded clauses, and more advanced structural arrangements. Conversely, AI-created texts have simpler, more linear, and formulaic structure, even though they are grammatically accurate with a fundamental coherence. The findings suggest that AI can be used to duplicate the superficial aspects of academic writing, but in more profound linguistic and rhetorical levels, it is much different. This paper underscores the relevance of human mental involvement in scholarly conversation and offers suggestions to the education practice and AI implementation.