Zero UI

Take the screen away and the interface does not disappear, it moves. Everything a layout was carrying, what is available, what state the system is in, what just happened, whether it worked, has to be re-expressed in language, sound, light, and timing, because those are the only channels left. That is the actual work of zero UI, and it is why the phrase is misleading: there is no less interface here, there is the same interface with the visual channel removed and every other channel doing overtime. The specimen above runs one household task with no display anywhere in it. What is drawn is a ring of light, a transcript of the audio, and a chime, and the transcript is on the page for the reader of this specimen rather than on the device.

Three constraints do most of the design work. There is no affordance surface, so nothing advertises what can be said or done and the system has to be discoverable through convention and onboarding rather than through a visible menu. There is no state display, so every confirmation has to be spoken, chimed, or lit, and a system that answers ambiguously is far worse here than on a screen, where a reader can simply look again. And there is no list, so anything with more than about three options has to be negotiated in dialogue, which is slow. Output is where cost and grace part company: a light ring that says listening is instant and unobtrusive, a spoken sentence is slow and unskippable, and the difference between a good voice system and a bad one is mostly the discipline to answer in four words.

The other half of the term is the part people quote and skip past. If the interface is presence, the room has to be watched, so a zero UI product is a sensing product, and the things it senses are the household, the body, and the schedule. The Calm Technology tradition, which is where the polite version of the idea comes from, argued for systems that inform through the periphery and demand attention only when they must. That standard is a real one and most voice assistants fail it, because the commercial pressure on an always-listening device points the other way.

The neighbour that has to be kept separate is multimodal interface, and the line is simple: a multimodal interface counts the screen among its channels and lets a reader tap, speak, or point at the same state, while zero UI removes the screen entirely, so the design work moves out of layout and into language, sound, and timing. Voice control is the channel most zero UI products are built on, the specific one that turns commands into sentences, and haptic feedback is how the same output problem gets solved when a device is on the body rather than across the room.