Web accessibility testing is most effective when it is treated as a repeatable release activity rather than a one-time scan. This guide provides a practical workflow for auditing keyboard navigation, focus states, headings, forms, color contrast, dialogs, landmarks, and ARIA, then turning the results into a regression checklist your development and QA teams can reuse.
Overview
A website accessibility audit combines automated checks, manual browser testing, and—when appropriate—testing with assistive technology. No single method catches every issue. Automated tools can identify many detectable patterns, but they may not understand whether a control's label is meaningful, whether a dialog behaves logically, or whether a keyboard user can complete a task without confusion.
Start with the user journeys that matter most: signing in, searching, submitting a form, purchasing an item, reading an article, changing account settings, or completing a support request. Test those journeys on representative pages and components instead of checking isolated markup only. A technically valid button can still create an inaccessible flow if focus disappears after activation or an error message is not communicated.
Use WCAG as a reference for requirements and success criteria, but translate each finding into a concrete implementation task. A useful issue report states the affected page or component, reproduction steps, expected behavior, observed behavior, impact, and suggested fix. Include a small screen recording or DOM snippet when it helps another developer reproduce the problem.
What to track
Maintain an accessibility checklist at the level of reusable components and critical user flows. The following areas cover common gaps in frontend implementations.
Keyboard navigation and focus
- Can a user reach every interactive control with the keyboard?
- Does the tab order follow a sensible reading and task order?
- Is the focused element visibly identifiable against its background?
- Can users operate menus, accordions, tabs, sliders, and custom controls without a mouse?
- After an action, does focus move to a logical location rather than vanish or jump unexpectedly?
Test with the Tab, Shift+Tab, Enter, Space, and arrow keys where the component uses them. Watch for positive tabindex values, click-only handlers, keyboard traps, and focus styles removed by CSS without an accessible replacement.
Structure, headings, and landmarks
- Is there one clear page purpose and a logical heading hierarchy?
- Do navigation, main content, complementary content, and footer regions use meaningful landmarks?
- Are landmark labels unique when multiple regions have the same role?
- Does the visual order match the DOM order?
Inspect the rendered accessibility tree as well as the source. CSS layout can make a page look correct while leaving screen-reader and keyboard users with an illogical sequence.
Forms and validation
- Does every input have a programmatically associated label?
- Are required fields and expected formats communicated before submission?
- Are related controls grouped with an appropriate fieldset and legend where needed?
- When validation fails, can users identify the field, understand the problem, and find the message?
- Are status messages and successful submissions announced without stealing focus?
Do not use placeholder text as the only label. Check autofill, error recovery, disabled states, and server-side validation as well as the initial empty form.
Contrast, text, and non-text content
Check text and controls in their normal, hover, focus, disabled, and error states. Test content over images, gradients, and changing backgrounds. Verify that information is not conveyed by color alone and that meaningful images have useful alternative text. Decorative images should not create noise for assistive technology.
Dialogs, dynamic content, and ARIA
For a modal dialog, confirm that it has an accessible name, receives focus when opened, keeps interaction inside the intended region, and returns focus to a sensible trigger when closed. Ensure background content is not accidentally available to users while the modal is active. For live updates, identify whether the change needs an announcement and choose the least disruptive implementation.
Use native HTML elements before adding ARIA. ARIA can describe or enhance a correctly structured control, but it does not automatically add keyboard behavior, focus management, state changes, or validation. A custom element with role="button" still needs appropriate keyboard handling and an exposed name.
Automated checks and manual evidence
Run an automated accessibility scanner against representative routes, component stories, and key states. Treat results as leads to investigate, not as proof that a page is accessible. Record the tool, browser, viewport, route, and test date so a later run can be compared fairly. Pair the scan with manual keyboard testing, zoom or text-resizing checks, and screen-reader spot checks for critical flows.
Cadence and checkpoints
Accessibility testing should match the way your product changes. A practical schedule is more useful than waiting for a large audit at the end of a project.
| Checkpoint | What to test | Output |
|---|---|---|
| During component development | Names, roles, states, keyboard behavior, focus, and error states | Component-level defects and reusable acceptance criteria |
| Before a feature merges | Critical flow, responsive layout, forms, dialogs, and dynamic updates | Issue list linked to the feature or pull request |
| Before release | Representative pages, navigation, authentication, and high-value tasks | Release accessibility checklist and known-risk notes |
| Monthly or quarterly review | Regression scan, browser and assistive-technology spot checks, and unresolved issues | Trend summary and prioritized remediation plan |
Keep a small test-page set that includes a content page, a dense data or settings page, a form with validation, a page containing a dialog or menu, and a responsive mobile layout. This set becomes a repeatable baseline after dependency upgrades, design-system changes, or framework migrations.
How to interpret changes
Compare findings by component, route, severity, and user flow rather than counting raw violations alone. A single issue in a shared button, modal, or form component may affect many pages and deserve priority over several isolated warnings.
Separate new regressions from findings that were already known. When an automated result increases, first check whether the scanner version, rules, route coverage, or rendered state changed. A higher count may reflect broader coverage rather than a genuine decline. Conversely, a stable count does not prove that keyboard or screen-reader behavior is stable.
Prioritize issues that block task completion, prevent keyboard access, hide essential information, create confusing focus movement, or affect repeated components. Then verify the fix in the original failing state and at nearby breakpoints. Add a regression test or component acceptance criterion when the defect could easily return.
For ARIA findings, inspect the accessible name, role, and state in the browser's accessibility tools. Confirm behavior, not just attributes. For visual findings, test real content and interaction states rather than a single static screenshot. For form findings, submit invalid data and follow the recovery path from the beginning.
When to revisit
Revisit this accessibility checklist whenever a shared component, navigation pattern, form system, CSS framework, rendering strategy, or browser-support assumption changes. Schedule a lighter regression review monthly or quarterly, depending on release frequency and product risk, and run a deeper audit after substantial redesigns or migrations.
Update your test plan when automated rules, browser behavior, assistive-technology support, or applicable accessibility guidance changes. Avoid treating an old tool configuration as a permanent definition of accessibility. Record the versions and environments used so future results remain interpretable.
For the next review, take these steps:
- Select the current critical user journeys and representative pages.
- Run automated checks on both default and important interactive states.
- Complete a keyboard pass from page load through task completion.
- Inspect headings, landmarks, labels, names, states, focus, contrast, and error recovery.
- Perform assistive-technology spot checks for the highest-risk flows.
- Compare results with the previous review and classify regressions separately from known issues.
- Assign owners, document verification steps, and add a regression check for every recurring defect.
A repeatable audit turns accessibility from a last-minute inspection into part of frontend quality. Keep the checklist close to your component and release workflows, and revisit it whenever the product or testing environment changes.