The numbers are striking, but the context matters more than the headline.
Turnitin scanned 3.1 million Australian university submissions between October 2025 and April 2026. More than half showed AI involvement of some kind. Ten percent had more than 80% of their content written by AI. For comparison, US students are doing this at nearly double the Australian rate: almost one in five US submissions crosses that 80% threshold.
At the same time, Anthropic published data showing Australia is now the world's top per-capita user of its Claude AI assistant, ranking first out of 121 countries. Australians use Claude at 6.4 times the rate their population would predict. That number jumped from 4.1 just two months earlier. Nearly 10% of Australian Claude conversations are tied directly to coursework, assignments, and study materials. These two datasets point to the same thing: AI is no longer a novelty for students. It is a daily tool.
The problem is that university credentials are used by employers as a proxy for capability. If a degree in law, nursing, finance, or logistics partly certifies that someone can structure an argument, analyse a problem, and communicate clearly, then the question of how much of that work was done by AI matters. Not because AI use is automatically wrong, but because no one currently knows how to measure what the student actually learned versus what the AI produced.
The detection tools universities use to answer that question are unreliable. Australian Catholic University flagged nearly 6,000 students for alleged AI misconduct in 2024. About 90% of those cases were AI-related. Roughly a quarter were later dismissed because the detection software was the only evidence. One nursing student had to wait six months to be cleared, had her transcript frozen, and missed out on a graduate placement as a result. She had not used AI. A US federal court ruled in early 2026 that a university cannot treat a detection score as proof and deny a student the right to respond.
Several major Australian universities have now disabled Turnitin's AI detection feature entirely. Curtin University switched it off in January 2026. The University of Queensland did the same in mid-2025. Some of Australia's most prestigious research universities still use it, but none without legal disclaimers and caveats.
The sector is moving toward a different approach: redesigning assessments so that AI cannot simply substitute for student thinking. This means oral exams, multi-stage assignments where students must explain their reasoning at each step, and grading that rewards the thinking process, not just the final document. Turnitin's own research found that more than 60% of educator feedback now prioritises transparency over detection: they want to know how a student used AI, not just whether they did.
For business operators, this has a practical implication. The graduates you hire from 2026 onwards have studied during a period when AI use was widespread but governance was inconsistent. Some will have genuinely learned to use AI as a thinking tool. Others will have learned to produce acceptable-looking output without understanding what is in it. The credential alone will not tell you which is which.
The smarter response is to test for judgment and reasoning in your own hiring and onboarding processes, rather than assuming the degree certifies it. Ask candidates to explain their thinking, not just deliver a polished output. That is, coincidentally, exactly what universities are now trying to teach.