If your everyday glasses contain a camera and an AI assistant, you are now carrying a prohibited device into a large share of proctored sessions — and in most hiring assessments, nobody will tell you that in advance. The rules are split. Testing bodies and certification programs have written explicit, capability-based bans that cover prescription frames. Technical hiring platforms mostly have not written anything, which does not mean the glasses are allowed; it means the question falls to a general prohibition on outside assistance and to whoever reviews your webcam footage afterward.
This article covers what the published rules say about smart glasses in proctored assessments and video interviews, what the review process can actually observe, and what to do before a session if you own a pair. Every rule quoted below comes from the testing body's or vendor's own candidate-facing documentation, read on 3 October 2026. The program you are testing under is the one that governs you.
The exam rules name them, and they are written by capability
The clearest current example is the SAT. College Board's testing rules, revised 29 September 2026, prohibit "electronic equipment, including phones, smartwatches, smart glasses or any eyewear that provides access to the internet or to any AI feature or service, fitness trackers, wearable technology (wearables), haptics, cameras, recording or listening devices, or any other type of electronic or communication device" (SAT testing rules, College Board). Violations carry "disciplinary measures and consequences, including score cancellation and other sanctions," up to "prohibiting you from taking future College Board tests including SAT, AP, or CLEP." Reporting from February 2026 placed the start of the smart-glasses prohibition in March of that year (Cybernews, 8 February 2026).
Read the wording carefully, because the drafting style is the point. The rule does not list banned products. It bans any eyewear that reaches the internet or an AI service. A brand list goes stale in a quarter; a capability test does not. It also makes no exception for corrective lenses, which matters because major AI glasses are sold with prescription options.
Professional certification follows the same pattern. CompTIA's online-proctored guidelines state that "Phones, tablets, smart devices, headphones, earbuds, watches—including smartwatches and smart glasses—must be removed and placed out of arm's reach," and that "Failure to adhere to any of these requirements may result in your exam being revoked, your results invalidated, and a possible permanent ban from future CompTIA exams" (CompTIA online-proctored exam guidelines). ProctorU's test-center item list is blunter still, prohibiting "Google and smart glasses (any glasses with electronics)" alongside watches and headphones (ProctorU, what to bring to an in-person proctoring session).
So in the exam world the answer is settled and the sanction is severe. If you are preparing for an AWS, Microsoft or CompTIA credential, the device question sits inside a broader set of environment rules that terminate more sessions than anything anyone would call cheating — what a remote-proctored certification exam records and can revoke later covers that layer in full.
Technical hiring platforms mostly do not mention them at all
Hiring assessment vendors have not matched this. HackerRank's pre-test candidate FAQ specifies the machine ("To take HackerRank Tests, you must use a PC or Laptop. Other devices(Mobile, iPad, etc) are not supported") and the browser, and tells candidates to "Use a web camera-enabled computer to take a proctored test." It does not enumerate prohibited wearables (HackerRank, frequently asked questions before the test). Mercer Mettl's test-taker guide likewise covers permissions, browser setup and the recommendation to avoid mobile devices, without a prohibited-items list (Mettl test taker guide).
Three things follow from that silence, and only one of them is good news.
The absence of a device list is not permission. What hiring platforms do publish is a prohibition on outside help, and that clause is written about the assistance rather than the hardware. CodeSignal states its intent is to let test-takers "access basic information about syntax…while still ensuring that test-takers are not receiving outside assistance" (CodeSignal, General Coding Assessment rules and setup). A rule phrased that way does not need to anticipate a product category to cover it.
Nobody warns you. A certification candidate clicks through an explicit agreement naming smart glasses. An assessment candidate gets an invitation email, a browser check and a webcam prompt. If you wear AI glasses daily, the first time the question arises may be in a reviewer's notes rather than in an instruction you could have followed.
The applicable rule may not be the platform's. Employers add their own instructions on top of the vendor's, and those travel by email rather than by documentation. Establishing what is permitted before you rely on anything is the same discipline that applies to AI assistance generally; how to verify the AI rules before a technical interview covers how to get an answer in writing instead of inferring one.
Live video interviews are the least governed case of all. A Zoom or Teams screening call usually has no published device policy, no room scan and no agreement to click. That is not a loophole so much as an absence of process: the interviewer's judgment, the recording, and the employer's general expectations are what remain.
What a review can actually observe
This is where the engineering detail cuts against the obvious assumption. Assessment telemetry does not enumerate your eyewear. HackerRank's proctoring captures "periodic snapshots during the test, which appear in the candidate's report for review," tracks tab focus, logs paste events, and runs "screenshot analysis, plagiarism detection, and webcam anomaly detection" (HackerRank, proctoring HackerRank tests). Every one of those signals is about your screen, your window focus or your face. None of them queries a Bluetooth peripheral or asks your operating system what is paired.
What does observe eyewear is a camera and a human. A webcam snapshot shows what is on your face. A room scan and a side camera show more; second-camera setups that frame your desk, hands and profile are now a default rather than a premium option on several platforms, as covered in what a phone used as a second camera actually sees.
And the hardware is unusually self-announcing. Meta states that whenever content is being captured, a white capture LED "blinks to let people know you're capturing content," that "The capture LED has no off switch," and that the glasses "disable the camera if they detect the LED was physically tampered with or destroyed" (Meta, AI glasses questions answered, 7 July 2026, updated 23 July 2026). A blinking indicator pointed at a webcam, inside a recording that a reviewer may open weeks later, is close to the worst possible evidentiary position: visible, timestamped, and hard to explain as something else.
It is worth being precise about the limits of that claim. A capture LED tells an observer that the camera is active. It does not tell anyone whether audio-only AI features were in use, and no published vendor documentation reviewed here describes detection of a wearable as a telemetry signal. Whether a given session is reviewed by a person at all, and how closely, is rarely disclosed.
The problem this creates for honest candidates
The uncomfortable case is not the person trying to cheat. It is the person whose prescription glasses happen to be a camera.
Capability-based rules do not care about intent, and the enforcement path does not start with a conversation. In the assessment context, an integrity signal typically arrives in a report a recruiter reads alongside your score, often without you being told it happened — the mechanics of that are set out in what happens when a coding assessment flags you. In certification, the sanction acts on the credential itself. In both, an unexplained device on your face during a recorded session is an adverse inference you will probably never get the chance to rebut.
The practical steps are unglamorous and effective:
- Have a non-electronic pair. If you depend on corrective lenses and your daily glasses are AI glasses, a plain backup pair is the single thing that removes the problem. Order it before you need it; exam bodies that ban the capability do not exempt prescriptions.
- Take them off before check-in, and put them out of frame. CompTIA's wording is "removed and placed out of arm's reach." Treat the stricter standard as the default when your own program is silent.
- Ask, in writing, when the session is a live interview. No published rule does not mean no rule. A one-line question to the recruiter costs nothing and creates a record.
- Do not rely on the indicator being unnoticed. The LED has no off switch, and blocking it disables the camera by design.
- Expect camera checks to get more direct. Interviewers are already improvising live verification; why interviewers ask you to pan the camera or wave a hand sorts those checks by what each one actually tests.
Where a desktop assistant does and does not fit
Control is a desktop-native AI interview assistant for interviews, assessments and screen-share workflows, available for Windows and macOS, with no Linux support (Control). It is a different category from a wearable, and the distinction is not cosmetic. A desktop assistant's behavior is a question about screen capture and what a meeting client transmits. A pair of glasses is a question about what a camera two feet away records, and no software property changes that.
It also does not fit where the rules exclude it. A remote-proctored certification exam prohibits assistance from any AI tool regardless of delivery mechanism, and that is a rule about the help, not the hardware. Where a live assistant is genuinely relevant is the interview or assessment where the employer permits it — and establishing that permission is step one, not an afterthought.
Where the rules are closed, the useful work happens beforehand. Control's free allowance is 5 messages and 2 minutes of voice (Control pricing), which is enough to rehearse answers against a rubric before a session you cannot retake.
The short version
In proctored exams and professional certification, smart glasses are prohibited, the rules are written by capability rather than by brand, prescription frames are not exempt, and the sanctions reach the credential. In technical hiring, almost nothing is written down, the governing clause is the general prohibition on outside assistance, and the enforcement mechanism is a human looking at recorded video rather than a device check. That combination is worse for you than an explicit ban would be, because it removes the warning without removing the consequence.
If you own AI glasses, buy a plain pair for interview days and take the smart ones off before check-in. If you want to know what a specific session permits, ask the recruiter and keep the reply.
Comparing tools? See how Control compares with Cluely, Interview Coder, Final Round AI, and others, or the best real-time AI interview copilots.
Share
Explore with AI
Continue exploring
Related guides
Guides · 8 min read
Does a Remote Interview Show Where You Actually Are?
Your IP is logged by the assessment platform, the video call, and the employer's fraud review. What each layer sees about your location, as of September 2026.
Guides · 9 min read
How to Get Your Interview Recording Deleted
Assessment platforms will not delete your interview recording on request. Who to send the request to, which deadline applies, and what deletion leaves behind.
Guides · 7 min read
What Happens When a Coding Assessment Flags You
HackerRank puts its AI plagiarism model at 85% precision and tells candidates nothing. Here is what an integrity flag actually does to your application.