Back to all posts
Quality7 min read·October 25, 2026

How to Check Software Accessibility Before Sign-Off

TB
ThynkBlox Team
Engineering

A software accessibility acceptance checklist should combine automated scans with hands-on tests of the tasks people need to complete. Before sign-off, check keyboard access, visible focus, labels, error recovery, zoom and assistive-technology use across those workflows. Treat a clean scanner report as evidence about the checks performed, not proof of full accessibility.

What should you agree before testing?

Start with the journeys that make the product useful: creating an account, finding information, submitting a form, completing a purchase or changing account settings. Choose those that apply to your product, including any essential administrator tasks.

For each journey, agree on the starting point, expected result and conditions to test. Include empty results, invalid entries, loading states and failed submissions. Define the devices, browsers and assistive technologies covered by acceptance.

If your agreement names an accessibility standard, record its version, level and scope. Use this checklist alongside that agreement; do not present it as a complete standard or a legal compliance assessment.

Connect these checks to your broader software testing and quality assurance plan, so accessibility has clear acceptance conditions.

What belongs in the acceptance checklist?

Use the table as a starting point. Replace general descriptions with observable results for your product.

AreaHands-on checkAcceptance condition
KeyboardComplete the journey without a mouseEvery required action is reachable and usable
FocusFollow the active control throughout the journeyFocus remains visible and moves to a useful place
LabelsInspect fields, buttons and grouped choicesNames and instructions explain their purpose
ErrorsSubmit missing or invalid informationErrors are discoverable, specific and recoverable
ZoomEnlarge content and repeat the journeyRequired content and actions remain available
Assistive technologyRepeat tasks with the agreed setupControls, states and outcomes are understandable
Automated scansScan relevant screens and statesFindings are reviewed, resolved or explicitly recorded

Mark each item as passed, failed or not tested. Record the reason for anything considered not applicable. An untested item should never disappear into a general “passed” status.

Can someone finish the workflow using a keyboard?

Begin at the entry point and put the mouse aside. Use Tab and Shift+Tab to move through controls, then the keys appropriate to each control to operate it.

Check that you can:

  • Reach every action needed to finish the task.
  • Open and operate menus, filters and date selectors.
  • Select options and submit forms.
  • Enter and leave dialogs without becoming stuck.
  • Reach actions that appear after another choice.
  • Use an alternative to dragging where dragging is required.

Do not improvise unfamiliar shortcuts until the interface works. If a control requires special keys, look for usable instructions and assess whether its behaviour matches the agreed interaction.

Follow the whole journey. Reaching the submit button is not enough if the next screen leaves the user unable to continue.

Is focus visible and does it move sensibly?

Focus is the location currently receiving keyboard input. During testing, require a clear visual indication of that location.

Check controls on light and dark backgrounds, inside scrolling panels and beneath sticky navigation. Look for focus indicators that are clipped, covered or difficult to distinguish from the surrounding design.

Then test transitions. After opening a dialog, check that focus moves into it. After closing it, check that focus returns to a useful place. When an action removes the focused item, require a deliberate next location.

Also check that moving focus does not unexpectedly submit a form or change the page. Record the exact point where you lose your place instead of reporting only “keyboard issue”.

Do labels explain the controls?

Read each form as though you have never seen it before. Require visible labels that remain available while entering information; do not accept placeholder text as the only instruction.

For each field, check:

  • What information is requested?
  • Is the field required or optional?
  • Are format instructions available before submission?
  • Are related choices grouped under a meaningful description?
  • Does the assistive-technology reading identify the same purpose?

Inspect icon-only buttons and repeated actions too. A row of buttons all announced as “Edit” may need more context to make the intended record clear.

Require clear names for consequential actions. “Delete saved address” gives a more specific acceptance target than an unexplained symbol.

Can someone recover from an error?

Deliberately leave required fields empty, enter invalid information and trigger a failed submission using an agreed test condition.

Require the interface to explain what went wrong and how to correct it. Check that the relevant field is identifiable without relying on colour alone.

For longer forms, test the error summary as well as the individual messages. Require a practical route from the summary to each affected field. Preserve valid entries where appropriate, and verify that correction does not force the user to restart the journey.

Repeat the error flow with assistive technology. Check whether the error is discoverable when it appears and whether the field's label, value and error can be understood together.

Finally, verify the outcome after retrying. The user should be able to distinguish a successful submission from another failure.

Does the journey still work when content is enlarged?

Increase browser zoom and text size using the settings agreed for acceptance. Record the settings so the check can be repeated.

Repeat the task rather than inspecting a static screenshot. Open menus, trigger validation, read help text and reach the final action.

Look for overlapping labels, clipped instructions, controls hidden outside panels and dialogs with unreachable buttons. Check that sticky headers and notices leave enough space to use the page.

For content that needs horizontal movement, such as a wide data table, assess whether its headings, cells and controls remain understandable. Keep that decision specific to the component; do not use it to excuse unrelated content disappearing.

How should assistive-technology testing work?

Choose the setup before sign-off and record the device, browser, assistive technology and relevant settings. Use a tester who knows how to operate that setup.

For a screen-reader check, repeat the agreed journeys and assess whether the user can:

  • Understand the page structure and find the main content.
  • Identify controls and their current values.
  • Understand expanded, selected and other relevant states.
  • Discover errors, progress updates and completion messages.
  • Complete the task without information available only visually.

Include other assistive technologies where they are part of the agreed scope. Do not turn one successful screen-reader session into a claim about every setup.

Where practical, include people who use the relevant assistive technology in everyday life. Give them a task and observe the result without coaching them through hidden workarounds.

What should the acceptance evidence look like?

Use a short record for each journey. An illustrative example could read:

Task: Change a saved delivery address.

Conditions: Keyboard-only navigation, enlarged content and the agreed screen-reader setup.

Result: The user opens the address form, understands its labels, corrects an invalid entry and confirms that the updated address was saved.

Evidence: Recorded setup, observed results, unresolved defects and the result of retesting fixes.

Avoid acceptance statements such as “accessibility looks good”. Require outcomes someone else can reproduce. Clear evidence also belongs in your software documentation, alongside the scope and known limitations.

What common mistakes should block sign-off?

Watch for these shortcuts:

  • Accepting only a scanner report. Request the hands-on results for the agreed journeys.
  • Testing only the initial page. Include dialogs, errors, loading and completion states.
  • Treating screenshots as interaction evidence. Require demonstrated task completion.
  • Accepting an undocumented workaround. Record the obstruction and assess the actual experience.
  • Closing a defect after an edit. Repeat the failing steps on the updated product.

Keep unresolved issues explicit. A release decision should identify affected tasks, the consequence for users and any agreed limits. Do not describe blocked or untested journeys as accessible.

Frequently asked questions

Does a clean automated scan mean we can sign off?

No. It supports only the checks that ran against the states examined. Require the agreed keyboard, focus, label, recovery, zoom and assistive-technology results as well.

Must every possible device combination be tested?

Define a practical acceptance scope before testing. Record the combinations examined and avoid extending the conclusion to combinations you have not assessed.

Is this checklist enough to claim compliance?

Use it to organise acceptance work. Any claim against a named standard needs evidence covering that standard's applicable requirements, beyond this introductory checklist.

When should accessibility checks happen?

Use them while reviewing designs and working features, then repeat the agreed acceptance journeys before sign-off. This gives each requirement a clear place in the delivery decision.

Ready to build?

Let's turn these ideas into your next product.

Start your project