Voice-control users operate by speaking the visible label
Aliases: Voice Control · Voice Access · speak the label
What it is
See “Submit” on a button, say “tap Submit.” Voice control, used as assistive technology, has the user speak the words on screen; the system matches that utterance to an operable object. It is not a hidden command language. The vocabulary is the visible labels.
Why it happens
These users can usually see the screen; hands or a pointer are unstable or unavailable. The strategy is look → read → speak, the inverse of a screen reader’s hear → jump. The matcher aligns the utterance with each control’s accessible name; the user builds the utterance from visible text. When the two agree, one hit; when they do not, the user repeats the visible label rather than guessing an internal field name.
The command space is the words currently on screen, not a global phrasebook such as “open settings.” Change the copy, and the speakable commands change. Voice control treats the visible UI as a spoken script, so users say whatever the label says.
Studying it
Turn on system voice control (Voice Control on macOS / iOS, Voice Access on Android). Speech only, no tapping. Have people submit an order, open a filter, delete an item, and record whether the first utterance is the visible text spoken as-is.
Independent variables: whether a visible label exists, whether it is unique, whether it is split from the internal name. Dependent variables: first-utterance hit rate, number of rephrasings, abandonments back to a gesture.
Do not show a “recommended commands” sheet first. The measurement is say-what-you-see, not whether they can memorize hidden commands.
Where it stops holding
Users who only hear (a reader at the same time) build utterances from announcements, not from visible words, and the claim weakens. System fallbacks such as “show numbers” are a recovery after a miss, not the primary path — the primary path remains speaking the visible label. Dialect, accent, and proper names lower recognition; that is the recognizer, not a reason users should stop reading labels. A control with no visible words gives this mechanism no material.
Applying it
- Every object that must be hit by voice needs a word on screen the user can speak.
- Write the visible copy of primary actions as the sentence a user would say, not an icon plus an internal name only developers know.
- How to check: turn on voice control and read the words on the button. If speaking the visible label hits it, the mechanism holds; if someone first has to ask “what is this called,” the UI is not yet a spoken script.