"I’m sure the more devious students know this and edit accordingly."
New data shows that two London universities expelled students for using artificial intelligence to cheat in the last academic year.
University College London and Imperial College London both removed students during the 2024-25 academic year, according to figures released under the Freedom of Information Act.
The disclosures come amid growing concern that universities are struggling to contain AI-assisted cheating.
Across the sector, more than 2,000 undergraduates at Russell Group universities were penalised for misusing generative AI tools in 2024-25. That compares with around 700 in 2023-24.
The rise has been reflected in expulsions elsewhere. The University of Glasgow and the University of Leeds also expelled students in 2025.
Some academics, however, argue the true scale of AI misuse is significantly higher than recorded cases suggest.
One professor from a Russell Group university said detection remains inconsistent:
“Formatting errors are usually also signs of copying and pasting.
“Quotation marks changing, different fonts, page numbers in referencing not adding up or just being wrong – those are the obvious signs.
“Sadly, I’m sure the more devious students know this and edit accordingly.”
The professor also described recurring patterns in suspected AI use, including AI-generated emails ending with “insert name here” and a submission incorrectly listing Michelangelo as the author of a 2017 book on the Kurds.
These examples, they said, reflect both overreliance on tools and gaps in student awareness of academic standards.
The Russell Group said the rise in cases may reflect improved reporting rather than purely increased cheating.
Hollie Chandler, director of policy at the group, said the number of students caught using AI “represents a very small proportion of the student population”.
However, seven of the 24 Russell Group universities admitted they do not record AI-related disciplinary investigations.
That comes despite a 2023 commitment from vice-chancellors to improve consistency in reporting across institutions.
The issue has also prompted wider changes in higher education.
Princeton University recently dropped a 133-year tradition of holding exams without invigilators. Concerns had emerged that its honour code may no longer be sufficient to guarantee academic integrity.
A survey by the Higher Education Policy Institute (HEPI) found 12% of British students admitted copying AI-generated text directly into assessments, up from 3% in 2023.
It also found 94% used AI tools in some form for assessed work, from revision support to drafting essays.
Rose Stephenson, director of policy and strategy at HEPI, said AI use had become widespread:
“At least some students are using AI to complete their assessments rather than engaging in the process.
“If this practice isn’t being identified, this generates a sense of unfairness with other students, who know it is happening, diminishes the value that the student using AI gains from the learning process and risks undermining the value of a degree.”
Loading ...