A Comparative and Diachronic Study of Linguistic Features in Human-Written and LLM-Generated Theses: Following the Linguistic Traces of Generative AI in Ca’ Foscari University Master’s Theses

The advent of Large Language Models (LLMs) and Generative Artificial Intelligence (GenAI) has enabled students from various educational levels to receive easily accessible assistance in completing challenging tasks. Although diverse types of human-authored academic texts were analyzed in terms of LLM-associated linguistic features, there is no previous study in the literature which focused on systematically comparing human- and AI-written master’s theses. The first aim of this thesis is to analyze the linguistic features of master’s theses written in Ca’ Foscari University between 2019 and 2024 and make a diachronic comparison between the theses of pre-ChatGPT (2019-2022) and ChatGPT era (2023-2024). The second aim of the thesis is to identify the linguistic qualities of AI-generated master’s theses and make a systematic comparison between human- and AI-written theses. For this reason, 161 theses were compiled from the thesis database, and their thematically matching counterparts were generated through gpt-5-mini-2025-08-07 in August 2026. The linguistic and statistical analysis yielded considerable differences in terms of average word length, proportion of content words, stop words, hedge words, passive voice, readability, and Biber features. The findings indicate distinctions between human-written master’s theses in the pre-ChatGPT and ChatGPT era, and between LLM-generated and human-authored master’s theses, which bridges a crucial research gap in the current literature.