The references turned up in macOS Tahoe 26.7, in code describing how AirPods with cameras would handle spatial audio, and were posted to the MacRumors forums by a member reading through it. The earbuds locate the physical screen you are watching, then compare that direction against what the motion sensors in the AirPods already estimate. Laptops, displays, phones, tablets and TVs are all in scope, and where several screens sit in front of you the audio narrows to the one you are actually looking at. The detection runs behind the scenes and may pause when you start moving.
Devices with Face ID already lean on those sensors to help track head movement. The code says that once the camera AirPods launch, the system prioritizes the camera data from the earbuds over Face ID input, which makes the buds the sensor Apple trusts most about where you are looking. The same code points at server-based visual reasoning, with camera data going to Apple servers to process visual intelligence requests. Rumors still put the launch at fall 2027, after the camera AirPods came off Apple's 2026 schedule, so this is a plan being written down well ahead of the hardware.