Does the recording light always come on?
It is supposed to. Ray-Ban Meta glasses carry a capture LED on the frame front that the device-either">wearer cannot switch off, and Meta pushed an update in July 2026 intended to detect a covered or damaged light and disable the camera when it finds one.
Whether that update works is contested. The Guardian reporting published on 19 August 2026 found a working market in disabling the light, with modification services saying their work still functions on current models after the update, and a journalist reporting success with an inexpensive sticker. Meta maintains that a covered or damaged LED disables the camera.
Treat that dispute as settled against you. Even if the safeguard held perfectly, the light is small, sits on one side of the frame, and competes with daylight and venue lighting. If the wearer is angled away, if you are in a crowd, or if you are simply not staring at their face, the signal exists and never reaches you. The absence of a visible light is not evidence of anything.
Can you tell smart glasses from ordinary glasses?
Less and less. Current models are built on standard eyewear frames from established brands and are designed specifically not to announce themselves, which is the commercial proposition. The camera housing sits in the corner of the frame front and reads as a design detail at conversational distance.
This is also why recordings turn up in places people assume are private. Living rooms, medical waiting rooms, doorstep transactions with a stranger. Identification by eye worked when the hardware looked strange, and it stopped working when the hardware became eyewear.
Can you just ask the person wearing them?
You can, and it is still the best single move available. Its limit is not about honesty.
A wearer often does not know. These devices have no screen, so the wearer's information about their own state arrives as a chime, a voice line, or a phone app they will not open mid-conversation. Someone can tell you truthfully that they are not recording and be wrong, because a session did not end when they thought it did. A sincere answer is not a verified one.
Does a Bluetooth scan detect them?
Partly, and the framing matters. Many devices broadcast Bluetooth Low Energy advertising packets that can be matched against known product signatures, which is how our own app works, and how the open-source detectors that have appeared this year work.
What a scan gives you is a floor, never a count. Devices rotate their broadcast address, so one pair can appear as several entries. Identification depends on a manually maintained list of signatures, so unbranded hardware stays invisible until somebody adds it. Advertising behaviour also varies by model and connection state, and NearLens has not verified a single connected-state rule across every model it tracks. And no scanner can read firmware, so detecting a model tells you what the hardware is, not what that unit currently does.
That is why any tool worth using should tell you nothing was detected rather than telling you nobody is nearby. Those are different statements and only the first one is true.
What actually helps
Assume the light is a bonus rather than a guarantee. Ask directly when it matters, and read the answer as sincere rather than verified. Use a scan as a floor.
And where the setting warrants it, a changing room, a clinic, a support meeting, push for a posted rule. A venue that states a policy at the door and backs its staff when someone invokes it does more work than every detection method on this page combined, because it shifts the question from what you can perceive to what is permitted.
