There is a quiet performance unfolding across parts of the higher education sector. On one level, institutions are signalling urgency, drafting policies, convening working groups and aligning themselves with regulatory expectations. On another, the day-to-day practice tells a more complicated story.
This is not about naming and shaming. It does not need to be. Staff and students already know how seriously their institutions are taking the misuse of generative AI, because they see it in what is done, not what is said.
Importantly, the narrative that academic integrity risk sits primarily with smaller or private providers does not always hold. In some cases, it is the larger, more established universities where the gap between policy ambition and operational follow-through is most visible.
With the Tertiary Education Quality and Standards Agency urging a rethink of academic integrity and assessment design, universities are, by any measure, busy. But activity should not be confused with effectiveness. There are several areas where that distinction is becoming increasingly apparent.
First, the basics. AI detection tools are widely available within platforms such as Turnitin. Yet in some institutions, these tools are not consistently activated across faculties or assessment types. In a setting where generative AI use is now mainstream, failing to enable detection by default is not a technical limitation. It is a decision.
Second, the burden on teaching staff. While policies evolve, expectations around monitoring are often less clearly articulated. Without consistent guidance, tools or institutional backing, academic staff are left to navigate AI misuse on their own. The result is predictable. Where enforcement is uneven, behaviours adjust accordingly. Students experiment, boundaries blur, and over time, what was once considered misconduct risks becoming normalised.
Third, the question of education. Generative AI is not going away, and nor should it be treated purely as a threat. The institutions that will navigate this shift most effectively are those that invest in teaching both staff and students how to use these tools responsibly. Avoidance is not a strategy. But nor is selective engagement, where responsible use is promoted without equally robust attention to misuse.
Fourth, assessment reform itself. There is broad agreement that traditional assessment models are under pressure. Yet not all redesign efforts are landing. In some cases, new assessments are no more resilient to AI-enabled shortcuts than the ones they replace. Reform, in this sense, risks becoming procedural rather than substantive.
Finally, the question of example. Students are acutely attuned to inconsistency. When they encounter AI-generated materials that are unacknowledged, or see staff using tools in ways that diverge from institutional guidance, the signal is clear. Integrity frameworks rely not just on rules, but on credibility.
None of this suggests that universities are ignoring the challenge. Far from it. The sector is engaged, and in many cases, deeply so. But there is a difference between responding to a problem and resolving it.
The institutions that will stand out in the next phase of this debate will not be those with the most comprehensive policies, but those where practice, culture and systems align. That requires more than compliance. It requires consistency, transparency and a willingness to confront uncomfortable gaps between intention and reality.
Because in the end, academic integrity in the age of generative AI is not something that can be performed. It has to be done.
