UK universities catch triple the AI-cheating cases in a year

Freedom of information requests to 131 UK universities, later analysed in an academic risk assessment and reported in the press, found nearly 7,000 proven cases of AI-assisted cheating in the 2023-24 academic year — a rate of about 5.1 cases per 1,000 students, roughly triple the 1.6 per 1,000 recorded the previous year.

The pattern varied sharply by institution. The University of Sheffield recorded 92 suspected AI-related cases in 2023-24, up from six the year before; Queen Mary University of London went from 10 suspected cases to 89; the University of Glasgow from 36 to 130. Queen’s University Belfast, by contrast, reported zero cases involving generative AI in either year — a gap the underlying analysis attributed at least partly to inconsistent detection and recording practices rather than genuinely different rates of misuse: more than a quarter of responding institutions said they had not tracked AI misuse as a distinct misconduct category in the earlier year at all.

The rise coincided with a fall in cases recorded as traditional plagiarism, from 19 to 15.2 per 1,000 students over the same period, though the analysis stopped short of asserting a direct substitution — AI-assisted misconduct and copy-paste plagiarism are typically flagged by different detection methods, so some of the shift plausibly reflects universities simply beginning to look for, and classify, a new category of case rather than a wholesale change in student behaviour.

The underlying data spans the 2023-24 academic year, but the fullest public accounting of it emerged later, once several rounds of freedom of information requests had been completed and compiled into the risk assessment cited here. The figures were widely read alongside continuing argument in UK higher education over whether AI-detection tools were reliable enough to support disciplinary findings, and whether assessment design needed to change faster than enforcement could keep pace with.