By Meena Jha and Amara Atif
Australian universities have moved fast on generative AI (GenAI). Policies, guidelines and statements on academic integrity are now widespread. But in classrooms, a more practical question is taking shape: what does using GenAI well actually look like?
Our recent research looked at this question through two postgraduate Information and Communications Technology (ICT) units in Australian universities. What we found is a growing gap between policy and practice. While institutions are setting expectations, educators are doing the heavy lifting of turning those expectations into something students can actually understand and use. And that’s not straightforward.
The policy-practice gap is real
Most universities now have guidance on GenAI. These documents explain what is allowed, what needs to be acknowledged and where the risks lie.
However these policies do not always address the everyday realities of teaching. Questions remain about how to design assessments that allow GenAI use while still making student learning visible, how to prevent over-reliance on these tools without banning them altogether, and how to help students critically question GenAI outputs rather than simply accepting them as accurate and trustworthy.
So, what happens? Educators figure it out themselves. This means GenAI integration is currently uneven some classrooms are doing thoughtful, scaffolded work, while others are still unsure where to start.
To explore what “good practice” might look like, we trialled two structured approaches.
Making GenAI visible: the GenAI usage template
In one unit, we introduced a simple but powerful idea: making GenAI use explicit. As part of their weekly tasks, students completed a short usage temp[late in which they identified the toll they used, explained how they used it, included evidence such as prompts or screenshots, and reflected on whether the output was useful, accurate or flawed
At first, students were unsure. Some used GenAI too much, others avoided it entirely.
But once expectations were clear, something shifted.
Students started thinking more carefully about their decisions. Instead of asking “Can I use GenAI?”, they began asking “How should I use it here?”. The template didn’t just track GenAI use it taught it.
In another unit, we took a different approach by embedding GenAI directly into a structured learning activity. Student began by using GenAI to generate an initial response, which they then critically evaluated before rewriting or improving it. They were also asked to reflect on what worked and what did not, and to share their insights with peers. This process repositioned GenAI as a starting point rather than an end point, encouraging students to engage more actively with the learning resources.
Through this approach students quickly realised that GenAI outputs can appear convincing, they are often too generic, may miss key details, and are sometimes simply incorrect.
The most valuable learning happened in the critique stage where students began spotting errors, questioning assumptions and refining ideas using their own knowledge.
In other words, they moved from using AI to thinking with and against it.
What we learned
Across both approaches, a few things became clear.
Students need structure not just permission: Simply allowing GenAI isn’t enough. Students need guidance on how to use it well. Templates, prompts and scaffolded activities help build that capability.
Transparency supports integrity: When students are asked to document their GenAI use, it normalises ethical behaviour. It shifts the focus from “policing” to “learning how to use AI responsibly”.
The real learning is in the critique: The biggest gains didn’t come from generating content they came from evaluating it. That’s where critical thinking, judgement and understanding develop.
But there are real challenges: This work is promising but it’s not easy.
Students start from very different places: Some students are confident GenAI users. Others are completely new to it. Designing for both groups at once is difficult leading to increased workloads.
Workload is increasing: Educators are redesigning assessments, creating new resources and interpreting grey areas around GenAI use. This adds to already heavy workloads. Are organisations ready to understand and see this increased workload?
There’s a risk of shallow learning: Without careful design, GenAI can encourage shortcuts. Students may focus on producing answers quickly rather than understanding the material.
Educators are doing more than teaching: One of the clearest insights from our research is this: educators are now acting as translators.
They are translating policy into practice, turning tools into meaningful learning opportunities, and reframing risks as teachable moments. This role is essential, but it can’t sit with individual educators alone.
So, what needs to happen next?
If GenAI is here to stay (and it is), then we need to move beyond ad hoc solutions. Here are four practical ways forward.
1. Make GenAI literacy a core skill
Students need to develop the ability to use GenAI effectively, to critically evaluate the quality and reliability of its outputs, and to understand the ethical implications associated with its use.
This isn’t an add-on it’s part of being a graduate in 2026 and beyond.
2. Give educators shared tools
Templates, examples and guidance shouldn’t be created from scratch in every subject. Institutions can support consistency by providing ready-to-use resources.
3. Rethink assessment design
Instead of asking how we can stop GenAI, the focus should shift towards how we design tasks where its use is both visible and meaningful, and how we can assess students’ thinking process rather than just the final outputs they produce.
4. Share the responsibility
GenAI literacy and academic integrity shouldn’t rest on individual educators alone. It requires support across teaching teams, learning designers, libraries and leadership.
Moving forward
GenAI has disrupted higher education, but it has also opened up new possibilities. Used well, it can support deeper learning, not replace it. It can help students develop the critical skills they’ll need in a GenAI-rich world. But this only happens when we design for it. Right now, educators are leading the way experimenting, adapting and figuring out what works. The next step is making sure they are supported to do this work, not just expected to carry it. Because the real challenge isn’t GenAI itself. It’s how we choose to teach with it.
Header image generated with Adobe Firefly
This article was originally published on EduResearch Matters. Read the original article.
The world is changing fast. We help you keep up.
We’ll send you 1 post, 3x per week.
