Prompted English: How Generative AI Is Reshaping Lexical Choice and Sentence Complexity in University Writing

Abstract

Generative artificial intelligence is becoming an active linguistic intermediary in university writing rather than functioning solely as a proofreading or information-retrieval tool. This study investigates how different levels of prompt-based AI assistance alter lexical selection and sentence complexity in English academic writing produced by university students.
A central research gap concerns the limited availability of controlled within-writer studies distinguishing linguistic changes caused by AI prompting from differences that naturally exist between human- and AI-authored texts.
The proposed study therefore compares independent student drafts, minimally AI-assisted revisions, and extensively prompted AI-assisted revisions produced from the same writing tasks and authors. Lexical behaviour is conceptualized through diversity, sophistication, density, academic-word concentration, and semantic replacement patterns, while syntactic change is examined through clause structure, dependency depth, subordination, and sentence-length variability. The study further proposes a Prompt Influence Index to estimate how strongly AI-mediated revision shifts a student's linguistic profile away from the original draft. Rather than treating more sophisticated vocabulary or longer sentences automatically as evidence of improved writing, the framework evaluates whether linguistic enrichment is accompanied by readability, stylistic consistency, and preservation of authorial voice.