Assessment is being reshaped by a simple reality: AI tools are now part of how students work, and many tasks that once showed understanding can be completed in seconds. While some students use these tools thoughtfully, others rely on them without really engaging with the work. At the same time, it’s becoming harder to keep students motivated, especially when tasks feel easy to automate and the purpose isn’t always clear.
This creates a real tension for educators. If assessment focuses only on final outputs, it becomes difficult to distinguish between genuine understanding and generated responses. The real challenge is not detecting AI use, but designing assessment that captures thinking, not just output.
In practice, this has required a shift in how we design assessment. Rather than focusing solely on what students produce, the emphasis needs to move toward how that work is developed, justified, and refined.
The first shift is to make thinking more visible within the task itself. Instead of relying on a final submission alone, students are asked to explain their decisions, outline their approach, and reflect on how their solution evolved. This doesn’t require major redesign, but it changes what is valued. The focus moves from output to reasoning.
The second shift is to design assessment around milestones rather than a single submission. Instead of relying on one high-stakes task, assessment is structured as a sequence of smaller stages that make student progress visible over time. This creates opportunities for feedback, iteration, and improvement, while also reducing reliance on a single final output. It changes how students engage with the work, as they are required to develop and refine their thinking across multiple points. The focus moves from submission to progress.
The third shift is how AI itself is positioned. Early responses often focused on restriction and detection, treating AI as a problem to manage. In practice, these tools are already part of professional workflows. A more sustainable approach is to assess how students use AI, how they interpret outputs, what they accept or reject, and how they justify their decisions. The focus moves from control to judgement.
These changes are not about eliminating AI from learning. They are about ensuring that learning remains observable. When students are required to explain their thinking, demonstrate progression, and exercise judgement, the integrity of assessment is strengthened regardless of the tools they use.
The challenge for educators is not to redesign everything at once, but to make small, deliberate shifts. Adding a reflective component, introducing a milestone, or asking students to justify key decisions can be enough to begin changing the nature of assessment.
In an AI-enabled environment, assessment can no longer rely on outputs alone. It needs to make thinking visible.
Alan Hatem is STEM Program Lead at UNSW College