UNSW's Can, Can't, Must assessment framework

The categories should not be read as a ranking from backward to advanced. An independent demonstration can be essential in a course that also makes sophisticated use of technology. Required AI use can be useful without becoming the appropriate condition for every task.

The actual instructions remain decisive

UNSW's student guidance still describes more detailed levels of assistance and directs students to course-specific expectations. Its School of Mathematics and Statistics, for example, states that generative AI assistance is not allowed unless specifically referred to in the assessment. A general partnership announcement cannot be treated as permission to disregard that instruction. [Student guidance]; [school policy].

Before using a tool, the student should establish what the task permits and how assistance must be acknowledged. If instructions appear inconsistent, ask the course convenor. Do not choose the broadest statement simply because it permits the most help.

This is a practical responsibility, not a test of whether a student is enthusiastic about AI.

Preparation should include evaluating work, not just producing it

A useful starting point is an assessment example in UNSW's guidance: students may be asked to critique an AI-generated draft. The university gives it as an example of AI-based assessment, not a requirement for every course. [Assessment examples].

For preparation, the important ability is to explain what makes an answer appropriate. Does it address the question? Is its evidence sound? Does a correction change the conclusion? Those are suggested learning questions, not an official UNSW admissions test or scoring framework.

A student who identifies a weakness should be able to explain its consequence. Merely labelling an answer biased, incomplete or inaccurate does not show that the student understands the subject. The explanation has to connect to the actual work.

Independent practice has a place alongside this. When the learning objective requires knowledge without assistance, the student needs opportunities to develop and demonstrate it. Families should not treat those sessions as obsolete simply because other work can be accelerated.

Ask for evidence of course design

When researching a programme, ask for an available assessment example and examine three things:

1. The contribution: What must the student decide or demonstrate beyond obtaining an output?

2. The check: Where does the course establish the knowledge needed to evaluate that output?

3. The feedback: Will the student understand why a decision was weak and have an opportunity to apply that learning?

These are questions for comparing course design, not a claim that UNSW has already answered them uniformly. If an assessment rewards a polished result but offers little way to examine the student's contribution, an AI partnership does not resolve that gap.

Confirm access for the actual intake separately. Then return to the education: how will this course help the student produce more demanding work and become more capable of questioning it? That is a stronger basis for choosing than the name of the software available.

Related Collegify pages: Undergraduate study planning and postgraduate study planning. Bring a course outline or assessment question to Collegify) when discussing preparation for Australian study.

DHS's fixed-admission rule is scheduled for 15 September 2026. What it could mean for F-1 students, travel, OPT and university planning.