Australian Universities Split on AI-Assisted Marking
Australian universities have split on AI-assisted marking, with some allowing limited AI under human judgment while UNSW, Melbourne, and Sydney ban generative marking.
PromptCrates Editorial
Staff Writer
Australian universities have split over whether staff may use generative AI to help mark assessments and write feedback, The Guardian reported on 28 September 2026 after a Sunday feature by Joe Hinchliffe. Western Sydney University, Newcastle, Deakin, RMIT, and Adelaide allow limited AI support with human final judgment, while UNSW, Melbourne, and Sydney ban generative AI for marking. Western Sydney senior lecturer Armin Alimardani warns of a “slop-cycle” and “verification drift” if overworked markers stop checking machine drafts.
Where campuses draw the line
The Guardian’s survey of institutional positions shows a qualified green light at several universities. Newcastle gives students an opt-out. Deakin says staff may use AI to improve efficiency and support assessment and administrative work but not to assign grades. Western Sydney’s spokesperson described a human-centred approach in which awarding marks, grades, and feedback remains the responsibility of academic staff even if AI tools support certain aspects. Adelaide and RMIT sit in the same limited-use camp according to the report.
Elite Group of Eight campuses take a harder line on marking even when they allow AI elsewhere in teaching. UNSW, the University of Melbourne, and the University of Sydney draw the line at generative AI for marking. Monash said AI is not used in marking at this time while it transforms assessment through a programmatic approach. Wollongong and James Cook answered similarly. Queensland’s University of Queensland stressed that academics must confirm learning outcomes and that grades reflect student attainment, while developing policies for the AI age. QUT’s “assessment uplift” language left some staff fearing fewer human marking hours for sessional academics who rely on that income.
The education debate sits beside other Australia AI governance stories on PromptCrates, including the Senate inquiry calling Altman and Amodei after rogue agents and U.S. state rules covered in Florida board AI rules for schools and colleges. Classroom tooling experiments such as Anthropic and CZI’s knowledge graph for classrooms show how quickly AI is entering pedagogy even as marking policy fractures.
Why Alimardani fears a slop-cycle
Alimardani, a law and technology lecturer at Western Sydney who studies AI closely, told The Guardian that colleagues asking whether AI could lighten grading felt “extremely concerning.” He worries higher education could collapse if students receive hollow feedback, employers discount degrees, and the economic case for enrolment weakens. The slop-cycle risk is mutual automation: students submit AI-written work and staff mark it with AI, producing credentials without meaningful human judgment.
He grants that AI could act as a “second pair of eyes” to catch inconsistencies and notes that many academics are overworked to breaking point with marking consuming large shares of time. But he says benefit claims are largely untested and that scale-up invites disaster. In his verification-drift scenario, if 100 academics receive an AI marking tool and 80 verify carefully, 20 will not; after a few drafts look fine, vigilance drops and hallucinations pass as grades. Courtroom scandals over AI citations, he argues, preview the reputational and fairness damage universities would face.
Student voice in the piece is blunt. QUT creative and professional writing student Alex Cameron said he would not pay for a degree marked by AI and would be “absolutely bloody incensed,” arguing that human writing deserves human readers and human feedback. Anonymous QUT academics linked vague uplift messaging to fear that sessional marking jobs will shrink. Integrity messaging that forbids student AI writing while allowing staff AI marking, they said, sends a contradictory signal.
What slow evaluation would require
If universities want limited AI assistance without collapsing trust, they need published rules students can audit: which tasks AI may draft, how humans certify final grades, opt-out paths like Newcastle’s, and sanctions when staff skip verification. Pilot studies should measure grade stability, feedback usefulness, and time saved before any whole-of-university rollout. Unions and sessional representatives need clarity on workload and pay when marking minutes fall.
Until those safeguards exist, the Australian split is itself the news. Some campuses treat AI as an efficiency aid under human responsibility; others treat generative marking as a bright line. Documented facts stay with The Guardian’s 28 September 2026 report: limited AI marking/feedback at Western Sydney, Newcastle (opt-out), Deakin (efficiency, not assigning grades), RMIT, and Adelaide; bans at UNSW, Melbourne, and Sydney; Alimardani’s slop-cycle and verification-drift warnings; Cameron’s refusal to pay for AI-marked degrees; staff concern over sessional jobs; several universities still formulating policy.
Regulators and quality agencies may eventually force a clearer national baseline if student complaints about AI-marked work rise. Until then, the market for international students—who pay premium fees and expect branded academic care—gives elite universities a commercial reason to keep humans in the grading loop even if domestic teaching budgets stay tight. Transparency about AI assistance in feedback letters would also help employers interpret transcripts.
For now, the Guardian package is a policy map, not a court ruling. Universities that allow limited AI marking still insist humans own final judgment; universities that ban generative marking still allow AI in other classroom tasks. The live risk Alimardani names is cultural: once verification feels optional, the human-centred policy becomes a slogan. Documented institutional names and quotes should stay tied to the Guardian article rather than to second-hand summaries.
Primary source: The Guardian on Australian universities and AI marking.
