The boundary is the point
The policies vary, but permission is generally conditional.
The University of Adelaide also allows some use. By contrast, the University of New South Wales, the University of Melbourne and the University of Sydney do not permit generative AI to mark work, even as they accept it in other parts of teaching.
Armin Alimardani, a law and technology lecturer at Western Sydney University, sees a potential benefit when AI acts as a “second pair of eyes”. He is working on models that identify inconsistencies in teachers’ marks. That promise lands in a system where, he says, many university staff are stretched and marking takes up a substantial share of their time.
But the benefits are mostly untested. Alimardani argues that claims about faster marking need to be examined slowly and carefully, and doubts AI can be scaled up for that purpose without serious consequences.
His concern is that human oversight will erode. Imagine 100 teachers are given an AI marking tool and told to check its scores: even if 80 do so, the other 20 may not. Alimardani calls the gradual loss of scrutiny “marking drift”. After several apparently sound results, an overloaded teacher may stop checking closely, allowing AI hallucinations to pass as facts.
The risk is not just an inaccurate grade. Alimardani warns that errors could damage universities’ reputations and leave students feeling unfairly treated, with doubts about the value of a degree. My read is that the policy debate is really about whether institutions can preserve meaningful human judgment while using a tool designed to reduce the work that judgment requires.
A message to students
The University of Monash says it does not currently use AI in assessment, though it is reshaping assessment as part of a broader program to bring AI into teaching and assessment. The universities of Wollongong and James Cook gave similar responses.
The University of Queensland says teachers must verify that students have met learning outcomes and that grades reflect their knowledge. It expects teachers to assess work and is developing rules to maintain that requirement. Queensland University of Technology, meanwhile, has begun a university-wide assessment improvement project and is gathering staff views on current practices, problems and opportunities.
The project’s language has unsettled some staff. One Queensland University of Technology lecturer joked that the university’s email sounded as though AI was telling employees it planned to take their jobs. A senior colleague, who asked not to be named, said teachers widely feared that the university might reduce manual marking where possible. For staff on temporary contracts, marking can be a main source of income in an already unstable field.
Students may see a contradiction too. Universities tell them that using AI on assignments raises academic-integrity concerns; applying AI to assess those same assignments could send a different signal. Alex Cameron, a creative and professional writing student at Queensland University of Technology, says the message would be that AI could do everything: write assignments and earn a degree.
Cameron edits the university’s Glass Magazine and says the value of university lies partly in learning alongside people with shared interests. He says he would not pay 1000 dollars for a degree if AI marked his work. Students want to write for people, receive feedback from people and know that people have read their work.
I think that is the question universities have not yet answered: if the human reader is part of what makes a degree worth paying for, how much marking can be handed to AI before the credential starts to mean something different?
Daily AI news
Every day we pick what actually matters in AI and explain it plainly — no hype, no filler. Subscribe if you want to follow where the industry is going.
Only what matters — every day
Follow on X