Screen-reader compatibility
Checks how assistive technology actually reads and operates an interface, including navigation modes and announcement clarity.
screen-reader-compatibility
See how it helpsWhen is it useful?
Use it when correct-looking markup still produces confusing speech, missing landmarks or unusable interactive controls.
What does it help you do?
- Checks actual browser and screen-reader pair behavior.
- Distinguishes reading navigation from focused widget interaction.
- Uses concise accessible names and reliable live announcements.
Try asking this
Verify a sortable table with VoiceOver and record what was actually spoken. Sort a column and move through rows in the tested browser-reader pair. Record the announced column, sort direction and row context rather than relying on markup alone.
Detailed documentation
Read the skill’s original instructions, examples and checks in their authored language.
Show less
Screen Reader Compatibility
Screen readers do not all interpret HTML and ARIA identically. How an interface is announced depends on the combination of operating system, browser engine, accessibility API, screen reader and navigation mode (browse/virtual cursor vs. focus/forms mode).
1. Primary Browser + Screen Reader Pairings
Target and evaluate markup against standard real-world pairings:
- macOS / iOS: VoiceOver + Safari (WebKit). Note: VoiceOver in Chrome has subtle differences, and WebKit removes list semantics (
role="list") from<ul style="list-style: none">unlessrole="list"is explicitly restored when list count semantics matter. - Windows:
- NVDA + Firefox or Chrome/Edge.
- JAWS + Chrome or Edge.
- Android: TalkBack + Chrome (explore-by-touch and swipe navigation).
2. Browse Mode (Virtual Cursor) vs. Focus / Application Mode
Desktop screen readers on Windows (NVDA, JAWS) operate in two primary modes:
- Browse / Virtual Cursor Mode: the user navigates headings (
H), landmarks (D/R), links (K), buttons (B), tables (T), lists (L) and form fields (F), and uses Arrow keys to read line-by-line or character-by-character through static text.- Implication: if you intercept
ArrowUp/ArrowDownon a regular container or putrole="application"on a page wrapper, you destroy the user's ability to read text and navigate by headings/landmarks!
- Implication: if you intercept
- Focus / Forms Mode: entered automatically when focus lands on a text
<input>,<select>,role="combobox",role="tablist",role="menu",role="slider"orrole="grid", passing keystrokes directly to the widget.- Rule: never apply
role="application"to an entire page or layout region; reserve it only for self-contained canvas/spreadsheet surfaces that manage 100% of keyboard navigation and announcements themselves.
- Rule: never apply
3. Concise, Natural Announcements (Speech Ergonomics)
Screen reader users listen to announcements linearly at high speech rates. Avoid verbose or redundant markup:
- Do not duplicate role words inside
aria-label:- Bad:
<button aria-label="Close button">-> announced as "Close button, button". - Bad:
<nav aria-label="Main navigation">-> announced as "Main navigation, navigation". - Good:
<button aria-label="Close">and<nav aria-label="Main">.
- Bad:
- Avoid giant concatenated accessible names: when a card or table row contains a title, status badge, timestamp, description and three action buttons, do not wrap the entire card in a single
<button>or<a href>that reads a 60-word paragraph as its accessible name. Instead, make the card heading the primary link (<a href="...">) and stretch its clickable hit area with a CSS::afterpseudo-element if full-card clickability is desired. - Hide purely decorative or redundant visual tokens:
- Add
aria-hidden="true"to decorative icons, visual divider glyphs (|,•), duplicate visual badges and background illustrations. - Ensure
aria-hidden="true"is never placed on a focusable element or its ancestor.
- Add
4. Tables, Lists & Description Relationships
- Data tables: ensure
<th>cells have explicitscope="col"orscope="row"so cell coordinates are announced when moving across columns and rows (Ctrl+Alt+Arrowin NVDA/JAWS,VO+Arrowin VoiceOver). - Supplementary descriptions (
aria-describedby): usearia-describedbyfor helper text, format hints ("At least 12 characters"), keyboard shortcut hints or field error messages so they are read after a short pause following the control's name and role without polluting the primary accessible name.
Included files 1
SKILL.md
