The Guardian reported on September 28 in Australia that some universities let staff use AI in marking and feedback, prompting questions from academics and students about the quality of assessment. The universities' own documents draw distinct lines. AI may help prepare feedback, contribute to an assessment process or be barred from marking altogether. A student still needs to know whether an academic read and judged the work.

The big change

  • What changed: Australian universities are spelling out where generative AI can enter assessment. Newcastle allows its use in marking under review, notice and opt-out conditions; UNSW tells markers, as a rule, not to use AI platforms for marking or grading.
  • Why it matters: A student can receive AI-assisted feedback without an AI system having final authority over a mark. The difference rests on what the academic actually checks and what the student is told before submitting work.
  • What to watch: Newcastle and Melbourne require notice and an opt-out route when staff use generative AI in assessment. Their public guidance does not show how often students receive that notice or how staff review the output.

Newcastle gives students a specific opt-out

The University of Newcastle's current policy on generative AI in teaching, learning and assessment says AI should not be used to mark student work unless the course marker or coordinator reviews and approves its output before sharing it with the student. Staff should tell students about use in the marking process before the submission deadline and explain how to opt out. Work submitted by a student who opts out should not be marked using generative AI. The policy covers coursework teaching, learning and assessment; its current version took effect on April 29, 2025.

Newcastle's staff FAQ, last updated February 10, 2025, adds that staff retain responsibility for the accuracy of marks. It describes educators using AI to draft feedback that they review and edit, while noting unanswered questions about effectiveness, privacy and intellectual property. Those are examples and cautions in guidance, rather than a published measure of how many assignments are marked this way.

Melbourne permits support, with approval and student choice

The University of Melbourne's staff guidance says staff may use secure university tools to support assessment or feedback after seeking their faculty associate dean's endorsement. They must tell students what tools will be used and how they can opt out. The guidance says staff remain responsible for evaluative judgments and feedback, must review AI output, and should not use a new generative AI tool alone to allocate marks or grades.

Melbourne also distinguishes its existing enterprise assessment tools, including automated multiple-choice marking, from proposed uses of new generative tools. Its opt-out guidance says a student should be able to rely solely on academic staff judgment for assessment and feedback.

By comparison, UNSW's staff teaching guidance says that, as a rule, markers must not use AI platforms for marking or grading. It separately tells course convenors to brief markers, including casual staff, on what AI use a particular course permits. Approval of tools for other teaching tasks does not overturn that marking instruction.

Western Sydney puts assessment duties on staff

Western Sydney University's current AI policy, effective July 30, 2024, applies to staff and students and calls for clear guidance wherever AI rules differ by setting or role. It does not itself set out a Newcastle-style marking opt-out or a detailed rule for AI-generated grades. Its assessment policy, effective January 1, 2026, puts the quality of feedback and consistent marking under academic staff and school responsibilities. A Western Sydney spokesperson told the Guardian that staff may use AI for certain aspects of assessment and feedback, but academic judgment and responsibility for marks, grades and feedback remain with teaching staff. That is an institutional statement reported by the newspaper, not detail written into the AI policy.

Deakin sets human-led principles

Deakin's AI Responsible Use Policy, effective March 12, 2026, applies to staff, students and associates. It calls for human-led decisions, evaluation of AI output and transparent communication when AI is used in teaching. It does not specify an opt-out for AI-assisted marking or define a permitted marking workflow. The Guardian reports Deakin's position as allowing AI to support assessment but not assign grades. That specific grade boundary comes from the newspaper's reporting; the publicly available policy sets broader principles.

The distinction matters to staff as well as students. In the Guardian's interviews, Western Sydney law and technology lecturer Armin Alimardani raised concern that a requirement to verify AI output could weaken in practice, and QUT academics described worries about casual marking work. Those are reported concerns, not observed rates of missed errors or evidence of job cuts. Neither the documents compared here nor those interviews establish that AI marking saves time or improves accuracy.

Australia's regulator, TEQSA, says it does not mandate a particular form of assessment; higher education providers must assess students' knowledge and understanding. Its generative AI resource hub shares institutional approaches for consideration, rather than a national rule authorizing or banning AI marking. For a student, the operative answer is therefore the university's current rule and the notice attached to the particular assessment.

Sources & further reading