Screen reader

Definition

A screen reader converts the accessibility tree into speech or braille output. Common examples include VoiceOver on Apple platforms, NVDA and JAWS on Windows, and TalkBack on Android.

People navigate, they do not listen linearly

The most common misconception is that a screen reader reads a page top to bottom. In practice people jump: to the next heading, to the next landmark, through a list of links, or through form controls. Reading everything in order is the slow fallback, not the normal mode.

What this means for markup

Structural navigation only works if the structure exists. Headings in the right order give a table of contents. Landmarks give regions to jump between. Labelled controls give a usable form list. A page built entirely from generic containers offers none of these, so the only option left is the slow linear read.

Output differs by product

Screen readers differ in what they announce, when they announce it, and how they handle ARIA. Testing in one is far better than testing in none, but a pattern that works in one is not proven everywhere. Prefer native elements, which have the widest and most consistent support.

Common failures

• Images whose alternative text repeats the adjacent caption.
• Link text that reads "click here" out of context.
• Headings chosen for size, producing a nonsensical outline.
• Form errors shown visually but never associated with the input.
• Content inserted dynamically with no announcement at all.

Review question

Turn off the display and complete one real task using only speech output. The question is not whether every word is announced, but whether the task can be finished without guessing.