Since ChatGPT entered our lives, many universities have relied on physical exclusion as an answer to assessment security in the context of artificial intelligence (AI).
If the problem is simply that we can’t tell whether AI or a student wrote an essay, then it seems we need only put the student in a situation where we can observe them under assessment conditions isolated from AI technology. In other words, we need only insist upon tasks such as invigilated exams and interactive orals and much of the AI assessment challenge goes away. However, AI glasses and other forms of wearable AI might soon undermine these in person assessment formats much as ChatGPT undermined written asynchronous tasks.
AI enabled smart glasses, such as Ray Ban Meta Glasses, can now read exam papers, listen to questions, and project AI-generated answers and assistance directly into the user’s line of sight. They often look like ordinary eyewear, they don’t produce obvious indications of use to observers, and for the user there’s no screen to glance at whilst no one is looking and no phone or notes to hide.
In our recent paper, On AI Glasses and Wearable AI in Assessment, Sue Sharpe, Phillip Dawson and I argue that in this way wearable AI can become ‘transparent’ both to the wearer (because, like all worn tools, it gets incorporated into ordinary perception) and to observers (because it produces no reliable external signal). You might be able to identify Ray Ban Meta Glasses, but, with new brands and form factors arriving constantly, you often can’t even tell which devices are AI-enabled.
This matters because physical exclusion, the logic underpinning many uses of exams and orals in the age of AI, only works if AI use is the kind of thing that can be cleanly separated from a student’s own activity, and the kind of thing that leaves some trace when it occurs. Wearable AI makes both of those assumptions shakier. It operates within the user’s perceptual field rather than alongside it, and it produces no characteristic signature, no gaze redirection, no posture of consultation, no second screen to glance at.
All that being said, it is important to note that wearable AI is not yet widespread, nor is it all that capable. However, seven million pairs of Ray-Ban Meta glasses sold last year, and Google, Apple and Samsung all have their own releases coming. What this suggests is that we need to start thinking now about what we’ll do when, or indeed if, the wind continues to blow in the current direction.
In 2022, after the release of ChatGPT, assessment found itself in chaos. No one wants to go back to those days but, if we wait until wearable AI is unavoidable, we are sure to scramble and we know what that looks like because we just lived through it. We have a chance to avoid repeating history, a window that is now open to think and plan for wearable AI entering the university. We should use that window before it closes on us.
Reference
Corbin, T., Sharpe, S., & Dawson, P. (2026). On AI glasses and wearable AI in assessment. Assessment & Evaluation in Higher Education, 1–17.
Share