These apps don't read a camera state or tap into anything resembling surveillance of the glasses themselves. They scan for Bluetooth Low Energy (BLE) advertisement packets, the short bursts of radio traffic a device sends out so nearby phones can find it, and compare what they see against a database of known manufacturer signatures (Ars Technica; VR.org).

The developer behind NoPeek, a free Android app, explained the technique on DEV Community: every BLE advertisement packet carries a company ID assigned by the Bluetooth SIG in its manufacturer-specific data field, and unlike a device's MAC address, that ID doesn't rotate (DEV Community). NoPeek checks for IDs tied to Luxottica, which makes Ray-Ban Meta glasses, along with Meta Platforms and Snap, and pairs the ID with a device name string so an iPhone's own Vision Pro doesn't trigger a false alert (DEV Community). Treat this as one developer's documented approach rather than proof that every smart-glasses product broadcasts an identifier nobody can turn off.

That approach has a hard ceiling built into it: a company ID identifies a manufacturer, not a specific product line. TechCrunch tested Nearby Glasses and found that adding Apple's company ID flooded the app with alerts from unrelated Apple devices nearby, not just headsets or glasses (TechCrunch). A broad Meta ID has the same problem, since it can just as easily flag a Quest headset as a pair of Ray-Ban Meta glasses. Looking for the hardware itself doesn't help either, because current Ray-Ban Meta and Snap Spectacles models are built to pass as ordinary eyewear, and even a strong Bluetooth reading can't tell you which face in a crowded room is wearing them.

These apps, in short, are repurposing a discovery signal the hardware was never designed to work as a public disclosure system. That's exactly why the output needs interpretation, not literal trust.

A detected signal doesn't mean detected glasses. Because identifiers work at the manufacturer level, an alert can point to a VR headset, an unrelated device from the same company, or a stray false trigger rather than eyewear (TechCrunch). Nearby Glasses' own developer, Yves Jeanrenaud, warns on GitHub that he's "no pro developer" and that detection "might sometimes just not work as expected," adding that false positives are likely (Ars Technica).

Silence proves even less. BLE advertising happens in short bursts, typically at power-on, during pairing, or when a charging case opens, because a radio that broadcasts constantly drains a battery already measured in single-digit hours (VR.org). A pair of glasses already bonded to its owner's phone, sitting quietly on someone's face, may not emit a single detectable packet the entire time it's near you (VR.org).

A wall, or even another person standing between the phone and the glasses, can block the signal outright (Ars Technica). On iOS specifically, background scanning slows down considerably, and Apple's own developer forums note that discovery can take minutes if a device isn't advertising frequently (Apple Developer Forums). If the app stays quiet, that's not evidence that nothing is nearby; it's evidence the app didn't catch anything this time.

Proximity readings are estimates, not a location fix. Zuckoff, one of the newer entrants, estimates distance from signal strength and assigns confidence bands such as "very close," "close," or "far" rather than a hard yes-or-no (Ars Technica). AntiZuck's own documentation is explicit that it doesn't identify the wearer, confirm recording, or report which direction a signal is coming from (VR.org). A "very close, high confidence" reading is a prompt to look around the room, not a way to single out the person standing next to you.

There's a forward-looking wrinkle worth knowing about. VR.org's reporting points out that BLE has long supported randomized, resolvable private addresses for other purposes, and that if a manufacturer applied similar randomization to the advertising payload itself, rather than just the MAC address, it could break every signature-based detector app overnight with no warning to users beforehand (VR.org). Nothing suggests that's happened yet. It's a plausible failure mode, not a current one.

Granting a detector app Bluetooth permission lets it listen for public radio broadcasts that nearby devices are already sending out, and nothing more. It can't see through walls, read what's on someone's phone, or peek at camera footage. AntiZuck and NoPeek both say their apps run entirely on-device, with no account, no analytics, and no camera, microphone, or precise-location access requested at all (VR.org; DEV Community). That's a reasonable baseline for the category, but it's specific to those two apps, not a guarantee covering every detector on the market.

If an app you're considering asks for more than Bluetooth, that's a personal privacy call worth pausing on, not proof the app is broken or malicious. A camera permission request from something whose entire job is listening for radio signals is, at minimum, worth a second look before tapping "allow."

1. Pick an app based on platform. Nearby Glasses costs $2.99 on both iOS and Android and has topped 100,000 downloads on the Play Store, making it the most established option (Ars Technica). AntiZuck is a one-time $2.99 purchase for iPhone and iPad covering Ray-Ban Meta glasses, other Meta-branded frames, Snap Spectacles, Amazon Echo Frames, and RayNeo (VR.org). Zuckoff launched free on iPhone in early August, with an optional $10-a-year plan that enables continuous background scanning (Ars Technica).
2. Grant only the Bluetooth permission the app asks for. Anything beyond that is worth questioning first, for the reasons above.
3. Start the scan. In Nearby Glasses specifically, PCMag's walkthrough shows users granting permissions, then tapping "Start Scanning" to begin listening for compatible signals (PCMag). Other apps in this category use their own interfaces and wording, so check each app's own instructions rather than assuming an identical flow.
4. Register your own glasses, if you own a pair. AntiZuck lets owners whitelist their own hardware so it stops triggering self-alerts, while still logging when it was last seen, a handy way to confirm the scanner is actually working (VR.org).
5. Read the confidence band, not just the ping. Weigh "far, low confidence" very differently from "very close, high confidence," but remember both are signal-strength estimates rather than verified positions (Ars Technica).

Don't confront anyone based on an alert alone. App developers explicitly discourage confrontation and disclaim liability for false positives or misread signals (Ars Technica). Treat a high-confidence, close-range alert as a reason to check your surroundings, not as grounds to accuse a specific person. In a sensitive setting, a locker room or a private event, that might mean asking venue staff about their policy rather than approaching whoever looks like they could be wearing glasses.

Built-in hardware indicators deserve the same skepticism as software alerts. Snap says its glasses include an LED and an audible tone when recording other people (TechCrunch). Meta's approach has drawn sharper criticism: Quintin called its recording LED "nearly useless" since it can be covered with something as simple as a sticker, and reporting from the Financial Times, cited by Ars Technica, suggests Meta may not activate the LED at all for an upcoming "super sensing" feature that would use its glasses' cameras and microphones more continuously (Ars Technica, citing the Financial Times). Neither indicator, present or absent, tells you as much as it's supposed to.

If something crosses the line into clear misconduct, report it through the channels built for that: venue staff, a platform's reporting tools, or law enforcement. A Bluetooth detector app was never built to settle that question, and its makers say as much.

These apps can genuinely surface a relevant signal often enough to be useful, which is part of why the category has grown from a single hobbyist project into several competing options in under a year (VR.org). But they can't reliably establish device type, distance, or whether anything is actually being recorded. An alert narrows where you look. It doesn't hand you proof.

Apple already runs unwanted-tracker detection at the operating-system level, a direct response to AirTags forcing an industry-wide standard rather than leaving the problem to independent developers (VR.org). VR.org's own analysis argues the underlying capability for smart glasses is much the same, though for now it sits with volunteer coders charging a few dollars rather than with manufacturers or regulators. EFF has called for a private right to sue over biometric privacy violations as facial-recognition features move closer to shipping in mainstream glasses (Ars Technica). Until something like that changes, using an app to detect recording glasses with Bluetooth is a useful habit for situational awareness, but venue rules, visible indicators, and actual evidence of misconduct still carry more weight than a radio signal ever will.