Every so often the discourse machine produces a headline that requires you to sit down. Here is one: plagiarism, dear reader, is down. Not up. Down. In the middle of the largest and most enthusiastically funded homework-cheating apparatus ever built for the general public, students are, on the whole, doing less of it.
The Numbers, Which Are Rude to the Narrative
Associate Professor Guy Curtis of the University of Western Australia's School of Psychological Science has been asking Australian university students to confess to plagiarism every five years since 2004, which is either extremely dedicated science or a mild form of professional masochism. In 2004, over 80 percent of students admitted to some form of it. In 2024 — surveying more than 2,100 students across six Australian universities, mostly at Western Sydney University — that number was 57 percent. Twenty years, one steady decline, and it happened on a timeline that includes both the smartphone and the chatbot that will write your essay for you while you make a sandwich.
Fifty-seven percent is still, to be clear, a lot of plagiarizing. Professor Curtis is not out here throwing a parade. But the trend line points the wrong way for anyone who'd built a tidy story about AI turning the modern classroom into a free-for-all.
Where the Robots Actually Show Up
Fourteen percent of students admitted to copying from AI without credit in 2024. That sounds like the headline number until you look closer: only 2 percent used AI copying as their only method of cheating. Everyone else was mixing it in with the traditional stuff — the copy-paste, the borrowed paragraph, the group project someone definitely didn't write alone. AI didn't invent a new kind of dishonest student. It just handed the existing ones a new tool, the way giving a teenager a nicer bike doesn't change whether they were going to skip school.
And here's the detail that should worry the ed-tech industry more than it worries anyone else: 88 percent of the students who copied verbatim from an AI tool knew perfectly well it counted as plagiarism. Not confused. Not misinformed about citation norms. Just doing it anyway, with full knowledge and, one presumes, full confidence they wouldn't get caught.
So Why Is Cheating Going Down, Exactly
Curtis credits two unglamorous, un-newsworthy causes: wider deployment of text-matching software, and — even more thrilling — better education about citation and referencing. Two decades of telling students how to properly footnote a source turns out to move the needle more than any single piece of alarmist AI coverage. Sources confirm this is not a satisfying answer for anyone who wanted a morality play, but it is apparently the correct one.
And yet — as if this were not enough to complicate the narrative — universities that deployed AI-detection tools showed roughly the same plagiarism rates as universities that didn't bother. The detectors, in other words, aren't deterring much of anything. They're mostly theater: expensive, imperfect theater that flags nervous international students and misses the kid who knew exactly what he was doing and did it anyway.
The Part Everyone Skips
The most inconvenient finding here is the 88 percent. The dominant assumption — the one baked into every op-ed about AI eroding academic integrity — is that students are being led astray by a confusing new technology they don't fully understand. Curtis's data says otherwise: most plagiarism, AI-assisted or not, is a knowing choice, not an accident of ignorance. Which means the fix was never going to be a better detector. It was always going to be a better reason not to cheat in the first place, and apparently reference-list literacy gets you further than panic ever did.
Even now, the alarm-industrial complex will find a way to be alarmed about this. But the data says what it says: the kids were already copying. They've just been given a shinier way to do it, and somehow they're doing it less.
Sources: University of Western Australia — Does AI mean more uni students are plagiarising their work?
