Google has filed a patent application for AR glasses that recognize a user's questions from a finger-pointing gesture and gaze direction. The application was filed on 9 April 2026 and published on 15 October. The system requires no voice commands or touches — simply pointing at an object is enough.
Two neural network modules work together. The first converts hand movement, finger position and gaze direction into a text query. The second finds the answer and displays it as a virtual label or line "attached" to the real object in the field of view. The glasses take into account the duration of the gesture, the distance to the finger and the context of surrounding objects.
An example from the description: when setting up a router, the user points at a port, and the glasses immediately show connection instructions right on the device. The system can also read text from labels and warn about risks — for example, if a cup of coffee is standing next to a hard drive. Additionally, the gesture can be combined with a short phrase such as "how do I connect these parts."
The patent mentions on-device processing to blur faces and documents before sending data to the cloud. This reduces privacy risks. However, a patent does not guarantee that a product will appear — Google regularly registers ideas, many of which never reach mass production.
The patent complements the Android XR direction and the Gemini integration. Earlier at I/O, Google demonstrated prototypes of glasses with advanced AI, but there is still no specific timeline for the release of a model with such a feature. Competitors — Meta and Apple — are also actively patenting similar solutions for AR glasses.
The patent describes a convenient way to interact without extra steps, but real-world accuracy and convenience in everyday use have not yet been tested.