The crisis facing our universities—that AI detection can no longer reliably establish whether submitted work reflects a student’s own knowledge—is not merely technical. It is part of a broader reckoning in education; including medical schools, where the line between a student’s reasoning and AI-generated output is increasingly difficult to police.
Our Higher Education Commission has admitted that existing AI-detection systems are “no longer reliably doing the job.” That exposes a weakness in traditional assessment: the assumption that a finished product proves the thinking behind it.
Better detection software will not solve this. A better algorithm may simply raise the cost of evasion; it does not restore trust. Universities must assess learning differently—observing and evaluating the process through which students develop, explain and defend their thinking.
That means more conversation, questioning, oral defense, practical demonstration and sustained interaction. Not because writing is unimportant, but because educators need to see the mind at work. University lecturers are already increasingly considering greater face-to-face engagement rather than relying so heavily on written papers.
Medical education faces a parallel question: how do they assess a future doctor’s judgment when a diagnostic note, treatment plan or patient summary may be generated or shaped by AI?
The concern goes beyond cheating. Educators are discussing de-skilling among experienced clinicians and “never-skilling” or “mis-skilling” among students who rely on AI before essential capabilities have developed. A student may produce a sophisticated diagnosis with AI assistance yet struggle to explain the reasoning behind it at the bedside. That gap is precisely what assessment must expose.
This is not an argument against AI.
AI is a profound game changer, transforming how we work, learn, communicate and make decisions. Whether we agree with it or not, it is changing our world, and we must embrace it intelligently—using its extraordinary capabilities without surrendering the human capabilities that make its use meaningful.
AI can be a tutor, simulator, drafting partner and source of instant feedback. But it cannot be the student.
For me, this is personal. I have spent more than 50 years traveling, working and learning across Australia, New Zealand, the US, Africa, West Asia and the Levant, Southeast Asia and Europe. Those experiences have given me a reservoir of knowledge, context and human encounters that shapes how I use AI.
When I ask AI a question, I do not approach its answer from a blank page. I can test what it tells me against what I have seen, experienced and learned over decades. I do not fear AI’s answers as much as I fear the absence of the inner library against which those answers are judged.
That distinction concerns me when I think about my grandchildren.
They are growing up in a world where AI can produce an apparently authoritative answer before they have acquired the knowledge and experience to question it. The danger is not simply that they may use AI to avoid homework. It is that they may outsource the very process through which judgment is formed.
They may learn to prompt before they learn to ponder.
An answer can be technically correct and still be wrong for the circumstances. Context, history, culture and human behaviour matter. Knowing what to ask—and recognising when a convincing answer does not make sense—often comes from accumulated experience, not information alone. Without that inner library, fluency with AI can masquerade as understanding.
AI can extend human capability enormously, but it cannot give a young person decades of lived experience. That experience still has to be acquired.
This is why the challenge facing universities and medical schools is larger than AI detection. It is whether education will continue to develop people who can think, rather than merely produce answers.
That requires more oral examinations, supervised work, iterative drafts, practical demonstrations, reflective explanation and problem-solving under observation. Portfolios that track revisions, oral defenses that probe reasoning and supervised tasks requiring live problem-solving can make thinking visible.
This is not a rejection of technology. It is about ensuring technology does not displace the intellectual development education is meant to cultivate.
The institutions that adapt will be those willing to rethink assessment: moving from policing outputs to observing process, from assuming competence to demonstrating it, and from asking whether a student produced an answer to asking whether the student can explain, challenge and own the thinking behind it.
The goal is not to make AI invisible.
It is to make human judgment visible.
The real educational challenge of the AI age is not whether our children learn to use AI. They must.
It is whether, in learning to use it, they also learn when to question it, when to challenge it—and when to think for themselves.
