Let's begin with the truth that makes us uncomfortable. Higher education isn't “broken” because of generative AI. The problem is no longer AI; it’s assessment. It has just shown us what we already knew: many of our tests were struggling to measure real learning. ChatGPT and other tools can now write essays, reports and even code in a matter of seconds. Instead of asking how to stop students from using AI, a better question is this: what are we really testing? Across the sector, there is growing recognition that traditional approaches, especially essays and unseen exams, are no longer reliable indicators of learning. Detection tools have emerged quickly, but their flaws are becoming clearer, especially regarding accuracy, bias and fairness.
So maybe the problem isn't how students act. It is the design of the assessment.
Moving beyond detection and surveillance
A common response to AI has been to double down on control: more detection software, stricter rules and tighter invigilation. But this approach comes with real risks. Discussions in the sector increasingly emphasise that surveillance-driven methodologies can erode trust and disproportionately affect specific student demographics, including international students and individuals utilising assistive technologies. In other words, trying to “protect” integrity may make unfairness worse without meaning to.
Detection is not very helpful for learning, which is more important. It works after the fact, focusing on law enforcement instead of teaching.
If we want assessment to remain credible, we need to shift from catching misconduct to designing it out.
What does AI-resilient assessment look like?
The good news is that we don't have to start over. A lot of what makes assessment "AI-resilient" is just good teaching that is done on purpose.
A few important ideas are coming out of both research and sector guidance:
1. Focus on authentic tasks
It's harder to use AI for assessments involving real-life situations. When students use what they know in real-life situations, organisations, or their own lives, generic AI outputs quickly fall short. For a long time, Advance HE has favoured authentic assessment to improve engagement and graduate outcomes. AI just makes the case stronger.
2. Design for process, not just product
Instead of relying on a single final submission, we can assess the learning journey. Drafts, reflections, peer feedback and short oral discussions all help make thinking visible.This aligns with the trend toward "assessment for learning" rather than "assessment of learning".
3. Make expectations explicit
Students are having a hard time right now because they don't know what will happen. What is okay to do with AI? What isn't? It is important to have clear rules, open conversations and clear standards. Students are more likely to act ethically when they know what is expected of them.
4. Treat AI as a learning tool
Banning AI outright is neither realistic nor educationally helpful. Instead, we can design tasks where students must use, critique and reflect on AI outputs. This shifts AI from a shortcut into a subject of critical thinking.
Why inclusion must be at the centre
There is another important dimension here that we cannot ignore: equity.
Assessment has never been neutral. It reflects particular assumptions about language, knowledge and ways of expressing understanding. AI risks amplifying these issues if we are not careful. For example, detection tools have been shown to yield higher false positives among students whose writing deviates from dominant norms. This raises serious questions about fairness.
Advance HE guidance consistently emphasises inclusive assessment design offering flexibility, clarity and multiple ways for students to demonstrate learning. In an AI-enabled world, this is no longer optional. It is essential. Inclusive design not only reduces barriers but also reduces the conditions that lead to academic misconduct in the first place.
A shift in mindset
Perhaps the biggest change required is not technical, but cultural.
We need to move:
- from suspicion to trust
- from rules to judgement
- from products to processes
- from standardisation to inclusion.
This is not about lowering standards. It is about aligning assessment with what we genuinely value: critical thinking, creativity, ethical reasoning and meaningful learning.
As many Advance HE conversations suggest, the future of assessment lies not in tighter control, but in better design and stronger relationships with students.
Where next?
Rethinking assessment design will take time. It requires investment in staff development, space for experimentation and institutional support.
But the direction of travel is clear. AI is not going away. The question is whether our assessments will evolve with it. If we get this right, we have an opportunity to design assessments that are not only more resilient to AI, but also more inclusive, engaging and educationally meaningful.
A question for you
What is one assessment in your own teaching that could be redesigned to focus more on process, authenticity or student voice in an AI-enabled world?
Dr Patrice Seuwou is an Associate Professor in Learning and Teaching with expertise in inclusive curriculum design, student partnership and digital pedagogy. His work focuses on operationalising equity and belonging through co-creation, Active Blended Learning and responsible use of Generative AI in higher education.