Screen Reader Design
Screen reader design creates meaningful structure, names, states, order, and feedback that assistive technology can present nonvisually.
Download Markdown (.md)Soroush’s Product Wiki · https://wiki.soroush.uk/topic/screen-reader-design/
01What it means
Screen readers use platform accessibility APIs derived from semantic HTML, accessible names, roles, properties, states, relationships, and live updates. Designing for them means specifying this nonvisual interface alongside the visual one, while preferring native semantics over custom ARIA.[1][2][3]
- Task + content structureOutline, landmarks, lists, tables, reading order
- Native semanticsLinks, buttons, forms, headings
- Names + relationshipsLabels, descriptions, ownership
- State + updatesExpanded, disabled, loading, error, live changes
- Accessibility APIScreen reader navigation and announcements
Prefer native semantics, specify dynamic announcements deliberately, and test real tasks with screen readers and users.
02Why it matters
Visual layout alone communicates hierarchy, grouping, status, and control purpose to many users. If those relationships are absent or misleading in code, screen-reader users may encounter unlabeled controls, noisy repetition, silent changes, or an order that does not match the task.
03When to use it
Include screen-reader behavior when designing forms, icons, tables, navigation, dialogs, validation, asynchronous updates, and custom widgets. Test common task paths with more than one relevant browser and screen-reader combination when risk justifies it.
04How it works
- Define a logical document outline, landmarks, lists, tables, labels, and reading order.
- Give controls concise accessible names that match visible language and expose state.
- Prefer native elements; use ARIA only to supply semantics HTML cannot provide.
- Specify what dynamic updates should announce and avoid excessive live-region noise.
- Keep hidden, disabled, loading, and error states understandable nonvisually.
- Inspect the accessibility tree and test real tasks with screen readers and users.
05A practical example
An icon-only mute button has the accessible name “Mute,” exposes its pressed state, and updates to “Unmute” consistently with the visible tooltip. Joining a call announces the new status once, while decorative participant animations remain hidden from the accessibility tree.
06Where it breaks
Screen readers differ by platform, settings, browser, and user skill. Passing one scripted test does not represent all users, and adding ARIA can make native semantics worse. Nonvisual access must be evaluated together with keyboard operation and clear content.
Common mistakes
- Using placeholder text as the only field label.
- Adding redundant ARIA roles or names to semantic elements.
- Announcing every visual update through an assertive live region.
- Hiding focused content with aria-hidden.
- Writing accessible names that do not match visible labels.
07Sources
- ARIA Authoring Practices GuideW3C Web Accessibility Initiative · Technical guidanceARIA Authoring Practices Guide
- Providing Accessible Names and DescriptionsW3C Web Accessibility Initiative · Technical guidanceProviding Accessible Names and Descriptions
- Introduction to Web AccessibilityW3C Web Accessibility Initiative · Institutional guidanceIntroduction to Web Accessibility