A University of Warwick-led analysis of almost 5,000 student-authored reports suggests that student writing has become more polished and formal since the introduction of ChatGPT in late 2022— but grades have remained stable.
Published in Computers and Education: Artificial Intelligence , the new study examines student reports submitted over a 10-year period and finds that the 'language' in students' writing has become more sophisticated, formal, and positive since 2022, coinciding with the widespread availability of generative AI (GenAI).
GenAI tools such as ChatGPT and Copilot are now widely used across higher education with a recent sector-wide survey showing that up to 88% of students report using ChatGPT for assessments .
This new analysis of 4,820 reports, containing 17 million words, is one of the largest of its kind. The experiment does not assess individual student's AI use but instead explores how writing has evolved at a cohort level during a period of rapid technological change.
It found that since 2022, writing sentiment has become more positive overall, regardless of the substantive content of the reports. This mirrors well-documented positivity tendencies in many GenAI systems, which are designed to produce polite, constructive-sounding responses.
Dr. Matthew Mak, Assistant Professor in Psychology, University of Warwick and first author said: "The tone of students' writing appears more positive, in line with ChatGPT's output, which is not inherently a good or bad thing, but it does raise concerns about the possibility of AI tools homogenising students' voices.
"There are also psychological studies showing that we tend to be less critical when we are in a positive mood; if students constantly receive GenAI output, it raises important questions about how these AI tools shape students' critical thinking in the long term."
The study also found significant increases in formality and range of vocabulary after ChatGPT's launch. These stylistic features would be expected to appear after many years of writing experience, making it unlikely this is a natural development in students' writing abilities nor does it indicate corresponding improvements in their underlying writing skills.
Additionally, some words frequently associated with AI-generated text, such as "delve" and "intricate", rose sharply in use until 2024 before plummeting in 2025, suggesting that students may have moderated their use—to make their writing read less AI-assisted.
To better understand these trends, the researchers also asked ChatGPT to rewrite reports submitted before ChatGPT was launched in 2022. These rewritten reports exhibited similar shifts in tone and style as in those submitted after ChatGPT's launch, providing additional evidence that the observed cohort-level changes are influenced by students' engagement with GenAI tools.
Importantly, despite these stylistic shifts, there was no corresponding changes in grades or examiner feedback. This may suggest that core academic skills — such as critical reasoning, interpretation, and argumentation — remain central to assessment and have, at least, not yet been overshadowed by changes in surface-level style brought about by ChatGPT.
Professor Lukasz Walasek, Department of Psychology, University of Warwick, author of the paper added: "Our findings highlight a transition in writing style that is likely happening across sectors. It is vital that institutions understand how tools like GenAI interact with learning and communication. This will help universities design assessments and guidance that support students to use these technologies responsibly and effectively."
The findings present opportunities for institutions to rethink assessment design, AI policy, and to support students in developing strong, authentic writing voices in an AI-rich world.
