# Accessibility Testing: Prove Each Design System Component Works for Everyone

> Learn how design teams check keyboard use, screen readers, zoom, and standards like WCAG on every shared component before any product uses it.

**In short:** Accessibility testing checks keyboard use, screen readers, zoom and standards like WCAG on every shared component before any product uses it.

From "Design systems", a free course by Venkat R. Murty: https://designervenkat.online/learn/design-systems/accessibility-testing

## What You Will Learn

This lesson shows how a design team tests accessibility for each shared component, not just for one screen. You will learn the three layers of testing, what to check on each component, how to record the results, and when and by whom the checks happen. A worked example walks through a modal dialog on a checkout screen, step by step.

- **Design system:** A shared set of parts, rules, and guidance that many product teams reuse.
- **Component:** One reusable part of a design system, such as a button, checkbox, dropdown, or dialog.
- **Accessibility:** Making a product usable by people with different abilities and in different situations.
- **Assistive technology:** Tools that help people use a screen, such as screen readers, screen magnifiers, and keyboards used instead of a mouse.
- **WCAG:** The Web Content Accessibility Guidelines, a widely used standard for accessible web content. It is made of testable rules called success criteria.
- **Keyboard navigation:** Moving through a page and using its controls with only the keyboard, usually with Tab, Enter, Space, Escape, and arrow keys.
- **Focus:** The marker that shows which element will respond to the next key press.
- **Screen reader:** Software that reads a page aloud, for example NVDA on Windows or VoiceOver on Apple devices.
- **Automated test:** A tool that scans a page and reports rule breaks without a person doing the check.
- **Manual test:** A check that a person does by hand, such as pressing keys or listening to a screen reader.
- **Contrast:** How much lighter or darker text is compared with its background. Low contrast makes text hard to read.

## What does accessibility testing mean in a design system?

Accessibility testing checks that each component works for people who use keyboards, screen readers, zoom, and other assistive tools. It also checks that the component meets standards such as WCAG. A single product team might test one screen once before launch. A design system works differently. Its tests are repeatable and are built into how components are made and released. That way, every product that uses a component inherits the same accessible behavior. When a team fixes a shared dropdown, every screen that uses it gets the fix.

> **Tip:** Test a component on its own first, then test it inside a real page. The same part can behave differently when it sits next to other content.

## What are the three layers of testing?

Think of accessibility testing as three layers. Each layer catches a different kind of problem, and none of them replaces the others. Skipping one layer leaves gaps that the other two cannot fill.

### Layer One: Automated Tools

Automated tools such as axe and Lighthouse scan the code behind a page and flag rule breaks. They are fast and they give the same result every time you run them. They are good at finding low contrast and missing labels. Their results are not the whole story, though. Sources disagree on how much they catch. Either way, a page can pass an automated scan and still be hard to use.

### Layer Two: Manual Keyboard Checks

In a manual keyboard check, a person uses only the keyboard and never touches the mouse. The New York State checklist asks testers to limit themselves to the Tab, Enter, and Space keys for this kind of test. The tester asks a few questions. Can I reach every control? Does focus move in an order that makes sense? Can I see where focus is at every step? Can I get out of the component again? Zoom belongs here too.

- **Reach:** Every button, link, field, and option can be reached with the keyboard.
- **Order:** Focus moves through the parts in an order that matches what the eye sees.
- **Visibility:** The focus marker is clear against the background in every state.

### Layer Three: Screen Reader Checks

A screen reader check asks what a blind or low-vision person hears. The tester listens for three things: the name of the control, its role (is it a button, a checkbox, a dialog?), and its state (is it checked, expanded, or disabled?). Test with at least two screen readers. NVDA is free on Windows, and VoiceOver is built into macOS. The same page can sound different in different screen reader and browser pairs, so write down which pair you used.

## What should you test on each component?

Each component needs its own checklist. The checklist links every check to a WCAG success criterion, so a failure points to a specific rule. A useful checklist covers six areas: contrast, target size, labels, keyboard support, focus management, and announcements.

- **Look and size:** Text and icons have enough contrast with their background in every state, including hover, error, and disabled. Clickable areas are big enough for people with limited hand control to hit.
- **Labels and names:** Every control has a clear name that a screen reader can read.
- **Keyboard and focus:** All actions work with the keyboard, focus is visible, focus follows a logical order, and focus never gets stuck.
- **Announcements:** Messages such as a saved confirmation or an error summary reach screen reader users without moving focus.

The checklist should also say how each item is tested. For example, a test might say press Tab from the top of the page and confirm focus lands on the component. A checklist item that cannot be tested in the design system alone should say so.

### How do you record results on the component page?

A test is only useful if the results are easy to find. Teams record each checklist item on the component's own page, so product teams can read it before they build. Digital.gov and the New York State system both publish per-component test pages for this reason.

1. **List every checklist item:** Write each item as a yes-or-no question, tied to its WCAG criterion.
2. **Run the tests:** Use the automated scan, the keyboard pass, and the screen reader pass for each item.
3. **Mark pass or fail:** Record the result for each item, along with the browser and screen reader pair you used.
4. **Note known limits:** Say what you could not test.
5. **Tie results to a version:** Store the results with the component version, so a later change does not hide an old failure.

## Worked Example: A Modal Dialog on a Checkout Screen

Imagine a checkout screen that opens a modal dialog to confirm a change of delivery address. The modal comes from the shared design system. Testing it here is part of testing the shared component, so the results apply to every product that uses it. The team runs the checks below in order. The expected behavior comes from the W3C dialog pattern and from the accessibility guides listed in the sources.

1. **Run an automated scan:** Open the modal and run axe on the screen. Check that the title and buttons have enough contrast and that the dialog and its close button have labels. Record the result, but do not treat a clean scan as a pass for the rest of the checks.
2. **Put the mouse away:** From the checkout page, move focus to the button that opens the modal and press Enter. Confirm that focus moves into the modal. The W3C pattern says that when a dialog opens, focus should move to an element inside it.
3. **Press Tab through the modal:** Press Tab and Shift+Tab many times. Confirm that focus cycles only through the modal's controls and never lands on the checkout page behind it. The W3C pattern says the tab sequence stays inside a modal.
4. **Press Escape:** Confirm that the modal closes. The W3C pattern lists Escape as the key that closes a dialog. If Escape does nothing, the component fails.
5. **Check where focus lands:** After the modal closes, focus should return to the button that opened it. The W3C pattern says focus normally returns to the element that invoked the dialog. Without this, a keyboard user loses their place on the page.
6. **Listen with VoiceOver:** Turn on VoiceOver, which is built into Apple devices, and open the modal again. Confirm that VoiceOver announces it as a dialog with its title. Repeat the test with NVDA on Windows if the team supports it.
7. **Record the outcome:** Mark each step as pass or fail. Note the browser and screen reader pair, and the limit that the automated scan cannot judge focus or announcements.

> **Tip:** Test the behavior with the keyboard, not just the code.

## When should you test in the component's life?

Accessibility testing is not a single event at the end. It happens at three moments, and each one catches different problems. Catching a problem early is cheaper than fixing it in every product.

### Testing at Each Stage

1. **Design stage:** Write annotations in the design file for focus order, contrast values, and the keys each control should respond to.
2. **During build:** The person building the component runs the automated checks and does a quick keyboard pass while the part takes shape. Problems found here are the least expensive to fix.
3. **Before release:** Run the full checklist, including the screen reader pass, before the component reaches product teams. Record the results on the component page.

### Who Owns Accessibility Testing

The exact titles vary by team, but the duties should be clear.

- **Accessibility lead:** Documents keyboard behavior, such as Tab order, activation keys, arrow-key behavior, and how focus enters and leaves a component.
- **Testing and QA:** Runs automated checks and manual WCAG and screen reader testing, then stores the evidence with the component version.
- **Design system lead:** Sets how severe a failure is, decides which failures block a release, reviews exceptions, and schedules a new check after changes to the component.

### Testing With Real People

Tools and checklists cannot confirm that a component is usable. A component can pass every check and still confuse a real person who relies on a screen reader.

> **Tip:** Ask testers to finish a real task, such as submitting a checkout form, rather than only checking one control. Tasks show problems that single-control checks miss.

## Check Your Understanding

Each question describes a new situation. Think about which layer of testing would catch the problem before you choose an answer.

**Quick check:** Your team's new dropdown passes an automated axe scan with zero errors. But a keyboard user cannot open it with Enter or move through its options with the arrow keys. Is the dropdown accessible, and what testing should have caught this?

A. Yes. Zero errors from an automated scan means the component is accessible, so no more testing is needed.
B. No. The team skipped the manual keyboard layer, which should have checked that Enter opens the list and arrow keys move through the options.
C. The problem is only visual, so a contrast check would have caught it.
D. No, but only a screen reader test on a phone would have found it.

**Answer:** B. Automated tools catch only part of the problems. Keyboard behavior needs a person pressing keys, so a manual keyboard check was the test that should have caught this.

**Quick check:** A shared form component shows a green message, Your changes are saved, after the user submits. Sighted users see it. A screen reader user hears nothing. The message passes the contrast check. Which test should the team add to its checklist for this component?

A. A larger target size test for the submit button.
B. A screen reader check that confirms the saved message is announced without moving focus.
C. A zoom test at 400 percent on the page header.
D. A second automated contrast scan of the footer.

**Answer:** B. The problem is about whether the message is announced. That is a screen reader check tied to status messages, and contrast or size checks cannot detect it.

## Sources

- [W3C WAI: Dialog (Modal) Pattern, ARIA Authoring Practices Guide](https://www.w3.org/WAI/ARIA/apg/patterns/dialog-modal/)
- [Utah Design System: Accessibility Checklist and Testing](https://designsystem.utah.gov/guidelinesStandards/accessibilityChecklistTesting)
- [U.S. Web Design System (digital.gov): Header accessibility tests](https://designsystem.digital.gov/components/header/accessibility-tests/)
- [U.S. Web Design System (digital.gov): Checkbox accessibility tests](https://designsystem.digital.gov/components/checkbox/accessibility-tests/)
- [U.S. Web Design System (digital.gov): Card accessibility tests](https://designsystem.digital.gov/components/card/accessibility-tests/)
- [New York State Design System: Accessibility Checklist for Tab](https://designsystem.ny.gov/components/tab/accessibility-tests/)
- [Select accessibility tests (U.S. Web Design System)](https://designsystem.digital.gov/components/select/accessibility-tests/)
- [Figr: Design System Accessibility Guide](https://figr.design/blog/design-system-accessibility)
- [CMS Design System: Step list component](https://design.cms.gov/components/step-list/)
- [University of Arizona: Testing toolkit](https://accessibility.arizona.edu/tools-checklists/testing-toolkit)
- [Pearly Quality: Manual vs Automated Accessibility Testing](https://pearlyquality.hu/articles/manual-vs-automated-accessibility-testing/)
- [Accessiway: Manual vs Automated Accessibility Testing](https://www.accessiway.com/blog/manual-vs-automated-accessibility-testing)
- [QA Skills: Accessibility Testing Automation Guide](https://qaskills.sh/blog/accessibility-testing-automation-guide)
- [A11yFlow: Automated vs Manual Accessibility Testing: The Data](https://www.a11yflow.dev/blog/automated-vs-manual-accessibility-testing)
- [Disability World: Screen Reader Testing Tools (NVDA, JAWS, VoiceOver)](https://www.disabilityworld.org/articles/accessibility-screen-reader-testing-tools-2026/)
- [DEV Community: AI Accessibility Testing Tools vs Manual WCAG Audits](https://dev.to/veduis/ai-accessibility-testing-tools-vs-manual-wcag-audits-what-actually-catches-what-48lf)
- [Accessibility.build: Accessible Dialog and Modal Guide](https://accessibility.build/guides/accessible-dialog)
- [Codeava: Accessible Modal Dialogs in React, Focus Management](https://www.codeava.com/blog/accessible-react-modal-focus-management)
- [UXPin: Keyboard Navigation Patterns for Complex Widgets](https://www.uxpin.com/studio/blog/keyboard-navigation-patterns-complex-widgets/)
- [TestMu AI: Accessibility Testing guide](https://www.testmuai.com/learning-hub/accessibility-testing/)
