Forensic stylometric analysis of AI-generated Text Compared Pakistani Student Academic Writing: Detecting Linguistic Fingerprints

Abstract

The rapid adoption of generative artificial intelligence (AI) in higher education has created new challenges for authorship verification and academic integrity, particularly in multilingual contexts such as Pakistan. This study investigates whether forensic stylometric analysis can distinguish AI-generated academic text from authentic academic writing produced by Pakistani university students through the identification of linguistic fingerprints. The study aims to compare the lexical, syntactic, discourse, stylistic, and authorship-related features of AI-generated and human-authored texts. A qualitative forensic stylometric approach was employed using a balanced corpus of 500 academic texts, comprising 250 authentic student writings and 250 AI-generated texts produced from identical writing prompts. Data were analyzed through systematic coding, categorization, thematic comparison, and interpretation of recurring linguistic patterns. The findings reveal consistent differences between AI-generated and human-authored writing, with authentic texts demonstrating greater lexical diversity, syntactic variability, discourse flexibility, and individualized stylistic markers, while AI-generated texts exhibit greater linguistic regularity and standardized patterns. The study concludes that forensic stylometric analysis provides a reliable framework for detecting linguistic fingerprints and strengthening authorship attribution in Pakistani higher education.