Definition: Voice control means operating a computer or phone with spoken commands – "Click Send" or "Scroll down", for example. On websites it only works when the visible text of a control is also part of its accessible name (WCAG 2.5.3) [1].
German: Sprachsteuerung · Also: speech recognition · Last reviewed: 27 September 2026
Who uses voice control
People who cannot use a mouse and keyboard, or can only do so in pain: after a stroke, with paralysis, muscular conditions, repetitive strain injury or an injury to the hands. The major systems have voice control built in – Voice Control on Mac and iPhone, Voice Access on Android and voice access on Windows – alongside specialist programs such as Dragon.
How it works
The software reads the page through the same interface as a screen reader, the accessibility tree. When someone says "Click Basket", it looks for an element whose name contains "Basket". If it finds none, the person can ask for numbers over every control ("Show numbers") and say "Click 14" – or bring up a grid and edge the pointer towards the target.
Why it fails
The most common mistake: visible text and accessible name do not match.
- A button shows "Send" but has
aria-label="Submit form". "Click Send" does not find it. - A link shows "More" but has an
aria-labelwith the article title. - An icon button with no visible text: nobody knows which word to say.
WCAG 2.5.3 Label in Name (Level A) therefore requires that when a control has visible text, its accessible name contains that text – ideally at the start [1][2].
Other obstacles are functions that only work by dragging or with complex gestures (WCAG 2.5.7), and single-letter keyboard shortcuts that spoken words trigger by accident (WCAG 2.1.4).
In code
<!-- The name does not match the text -->
<button aria-label="Submit form">
Send
</button><!-- The visible text is the name -->
<button>Send</button>How Reviseberg handles it
Reviseberg uses axe-core to check that links, buttons and form fields have a name at all – with no name, voice control finds nothing either. Whether the visible text is part of that name, and whether commands work in practice, is for a person to judge: a voice control agent is not part of Reviseberg today.
Related terms
Accessible name · Assistive technology · Target size · ARIA · Form label