Since February 2026, several AI glasses detection apps—including Nearby Glasses—have appeared, but what they pick up is the presence of a device, not its recording status. On August 21, 2026, Ars Technica reported testing an iPhone app called "Zuckoff," which detected a pair of Ray-Ban Meta glasses and correctly identified the model. However, what Zuckoff actually captured was a Bluetooth Low Energy (BLE) advertising signal. The app itself explicitly states that detection does not mean the device is recording, and that it cannot determine whether recording is underway.
This isn't merely a matter of detection accuracy. What people nearby actually need isn't a candidate signal suggesting "a device that looks like AI glasses is nearby"—they need an independent way to verify whether the act of filming and saving is happening at that moment. Current detection apps aren't reading a standardized signal that manufacturers have published specifically for recording status. Instead, they infer the presence of a device based on BLE advertisements broadcast for connection and discovery purposes.
Inferring Device Presence from BLE Advertisements
Zuckoff cross-references manufacturer signatures, service UUIDs, and advertised names contained in Bluetooth broadcasts. Results come with a confidence level—possible, likely, or strong—along with an indication of which clues matched. It also shows a rough proximity range based on signal strength, but by design it cannot indicate direction, and it doesn't display an arrow pointing toward the distance.
The same approach appears in Nearby Glasses, which emerged after February 2026. Published on GitHub, this app uses a heuristic based on company IDs, advertised names, and service UUIDs found in BLE broadcasts. Ars Technica's hands-on testing showed that such inference can work in practice in some situations. However, a successful detection is not proof that recording status can actually be read.
There is value in an app showing its reasoning. Users can treat the result not as "confirmed fact" but as a clue for identifying nearby devices. Even so, current BLE advertisements provide no basis for reinterpreting the meaning of these signals as evidence of an actual filming act.
The Gap Between Presence Detection and Recording Status
Zuckoff explains that once glasses are paired, they may weaken or stop their Bluetooth broadcasts. This means that a failure to detect the device doesn't prove recording isn't happening. Conversely, detection doesn't necessarily mean recording is underway. There remains an unfilled gap of information between candidate presence and actual recording status.
Nearby Glasses also warns of both false positives and missed detections. This is because company IDs such as Meta's can also be used by other products, including VR headsets. BLE broadcasts are also affected by obstacles, the human body, device orientation, transmission state, and OS background restrictions. A name appearing on screen cannot be extended into proof of an action actually taking place nearby.
None of this means detection apps are useless. They could plausibly help facilities enforce no-glasses policies or serve as a conversation starter. But if you expect such an app to serve as a tool for confirming whether recording is happening, you end up reading answers into it that the app never actually provides.
How Notification Is Designed for Saved Recording vs. AI Visual Use
Meta distinguishes between "Active Capture"—taking photos and videos meant to be saved—and "AI Features," which involve real-time visual understanding without saving. On Ray-Ban Meta, a capture LED lights up when taking photos or videos intended to be saved. According to Meta's FAQ, this LED blinks continuously during video recording, and there is no switch to turn it off. From the second generation onward, the glasses reportedly disable the camera if the LED is detected as being obstructed, strengthening detection of physical tampering or destruction as well.
According to a Meta document from July 2025, this LED measures 2mm, larger than the 0.9mm LED on the original Ray-Ban Stories. Meanwhile, the LED is not used when AI features are engaged in visual understanding. Meta explains that keeping the LED on at all times would cause people to grow accustomed to the notification, weakening its effectiveness as a signal for saved recordings, and cites other protective measures such as removing identifying information instead.
Breaking down what's visible to bystanders by device state makes the notification gap clear.
| Glasses State | Outward Signal/Protection Meta Describes | What Detection Apps Can Read |
|---|---|---|
| Saved photo/video capture | Capture LED lights up/blinks | Candidate device presence, if BLE broadcast is present |
| AI-driven visual understanding | No LED used; identifying information removed, among other measures | Candidate device presence, if BLE broadcast is present |
| Powered on / after pairing | Device behavior separate from capture state | Broadcast may be strong or may go quiet |
A visible LED signals saved capture to people nearby, but detection apps infer device presence from BLE signals alone. No cross-manufacturer, standardized signal has been published that would let a smartphone distinguish whether AI processing is occurring, whether it has proceeded to saving, or whether it has gone further to sharing or distribution. There is no technical link between Meta's protective measures and third-party apps.
The Precedent of DULT and Cross-OS Notifications
In response to the misuse of Bluetooth tracking devices, Apple and Google jointly developed "Detecting Unwanted Location Trackers (DULT)." iOS and Android issue warnings not to the device owner but to the person being tracked, providing steps to identify the device, locate it by sound, and disable it. Because AI glasses and tracking devices differ in purpose and threat model, this framework cannot simply be transplanted as-is.
Even so, DULT stands as a precedent for delivering a machine-readable, cross-OS notification to the affected party. Ars Technica reported that Zuckoff developer Pawel Szydlowski proposed that a rule requiring devices to broadcast a specific Bluetooth service identifier while recording could improve detection apps. This is neither an industry standard nor a plan Meta has announced it will adopt.
Even if such a status signal were standardized, design challenges would remain. If the externally broadcast identifier were persistent, a mechanism meant to verify recording status could become a new means of tracking the wearer. Separating recording-status verification from device/wearer tracking would need to be a precondition for any such specification.
Returning Self-Protection Apps to the Scope of Inference
At facilities that prohibit AI glasses, detection apps could serve as a supplementary verification tool. But an app's display alone cannot determine whether recording occurred or establish a rule violation. Current apps are tools for indicating "a plausible device" based on signals—not instruments that prove an action took place.
This also has implications for how facilities frame their own rules. Whether a rule bans "wearing or bringing in" the device, bans "using camera or AI visual features," or bans "recording that involves saving," the information needed to verify compliance differs. BLE detection can mainly assist with the first type of rule; the latter two cannot be confirmed by an app alone. The developer of Nearby Glasses has also asked users not to blame others or act hastily based on a warning. Detection results should not be reinterpreted as evidence of the act of recording.
Both efforts to improve LED visibility and BLE-based presence detection represent progress in addressing anxiety around recording. But the state that lies between them—whether filming with intent to save is happening right now—remains something bystanders cannot independently verify. Going forward, the deciding factor won't be detection accuracy rates, but whether manufacturers and OS makers can specify a way to make recording status verifiable without increasing tracking risk.