Wearable AI in exams refers to the use of smart glasses, discreet displays, or similar devices to capture exam content, send it to an AI model, and receive assistance back during an assessment. The core issue is not the novelty of the hardware; it is whether the assessment can still evidence the candidate’s own performance when hidden external help is available.

This matters because many exam security regimes are designed around obvious devices such as phones. Wearables can make unauthorised assistance harder to spot, which shifts pressure onto invigilation, device rules, and assessment design. It also raises practical questions about accessibility, legitimate assistive use, and proportionality.

- **Authenticity**: whether the work really reflects the candidate’s own knowledge and skill. - **Security design**: the extent to which rules, room checks, and invigilation can prevent or detect hidden assistance. - **Concealment**: how discreetly a device can be worn or used. - **Feasibility vs prevalence**: a demonstration can show that something can work without showing how often it is used in live exams.

The source set strongly suggests that wearable AI is technically feasible in at least some exam contexts. A project blog demonstrates an AI-assisted workflow using Rokid smart glasses, a phone, and cloud inference, and reports a score of 92.5 out of 100 in a computer communication networks final exam, compared with a student average of 72. The same source also notes practical constraints such as camera quality, battery life, power consumption, and connection stability. There is also broad practical agreement that this is an assessment design issue, not just a misconduct anecdote. If the only defence is spotting a hidden wearable, the assessment may be over-reliant on invigilator detection.

What remains unsettled is the scale of the risk. The available evidence supports a credible threat signal, but not prevalence, real-world detection rates, or the best countermeasures across different exam settings. The open question is how often wearable AI is used in live exams and which controls are effective without creating unfair or unworkable restrictions.

- Hidden external assistance may undermine authenticity and weaken score validity. - Security assumptions focused only on phones may miss camera-equipped wearables. - Overly broad restrictions may create accessibility problems or disproportionate controls. - Heavy reliance on detection can shift risk onto invigilators rather than assessment design.

Assessment teams may want to ask: - What device assumptions are built into exam rules and search procedures? - Are controls written for phones only, or for any camera-equipped wearable? - What evidence exists on detection reliability in realistic conditions? - How would controls affect accessibility and legitimate assistive use? - Does the task itself need redesign to reduce the value of hidden external help? The strongest next step is to treat wearable AI as part of broader exam authenticity and security planning, rather than as a niche gadget problem.

- Demonstration article showing an AI-assisted exam workflow with smart glasses and reporting a high score in a networking exam context.

Market signals suggest that smart glasses and other wearable devices are becoming more capable, but the cited source is a technical demonstration rather than independently validated evidence of widespread exam misuse. That means vendor and maker claims should be treated as signals of what is becoming possible, not proof of operational impact at scale.

### Can wearable AI be used to cheat in exams? The cited demonstration suggests it can be used in some conditions: a wearable device captured exam content, passed it to an AI model, and displayed help back to the user. That shows feasibility, not how common the practice is. ### Why does this matter if exams already ban phones? Because a discreet wearable may be harder to notice than a phone. That changes the security problem from device banning alone to broader assessment design, invigilation, and rules about wearables. ### Does wearable AI automatically make an exam invalid? Not automatically. The risk depends on the assessment design, the controls in place, and whether the exam still measures the intended knowledge or skill.

Tim Burnett (Admin)

Test Community Network. "Wearable AI in exams." TCN AI & Assessment Wiki. Last reviewed 2026-04-21. https://www.testcommunity.network/wiki/wearable-ai-in-exams.html

- Smart-glasses demonstration article describing an AI-assisted exam answering workflow and noting practical constraints.