Automated checks find certain accessibility problems quickly and consistently. Manual testing checks how the experience behaves when a person tries to use it. A useful review combines both and records what was actually tested.
The W3C’s evaluation guidance explains that some checks require human judgment. A clean automated report is therefore not a complete accessibility assessment.
What each method can tell you
| Area | Automated checks can help | A person must also inspect |
|---|---|---|
| Images | Flag images missing an alternative-text attribute. | Whether the text communicates the image’s purpose, or the image should be decorative. |
| Forms | Identify many missing labels or invalid attributes. | Whether instructions, errors and success messages make sense during submission. |
| Navigation | Find some structural or semantic problems. | Keyboard order, visible focus, menus and the ability to escape an overlay. |
| Dynamic content | Inspect the rendered state captured by the scan. | Other states, focus changes and whether important updates are announced. |
| Reading | Flag some contrast and document-structure issues. | Zoom, reflow, understandable content and the practical reading order. |
Test a journey, not only a homepage
Choose the tasks that matter: find a service, select a product, request a quote, change a cart or contact support. List the pages and states in each journey. A scanner visiting the initial URL will not necessarily open every menu, reveal every error or complete a multi-step form.
Use the keyboard to move through controls, activate them and return from dialogs. Check that focus stays visible and does not become trapped. Then inspect representative tasks with a screen reader and at increased zoom. Record the browser, assistive technology and the state being tested so someone else can reproduce the finding.
A form example
Imagine an email field with a proper label. An automated check may accept the label. Submit an invalid address: if the error appears only as a red border, the actual recovery experience still needs work. The fix might require an explanation associated with the input, appropriate focus handling and a clear way to try again.
Retest the error and the successful submission after the repair. Confirm the request reaches its intended destination as a separate delivery check. Accessibility and message delivery are both necessary for a useful inquiry flow.
Agree the review boundary
Before commissioning work, agree the templates, states, browsers and assistive technologies to cover, the standard being assessed and the format of the findings. Separate changes your developer can make from problems in an external plugin or hosted checkout.
Ask for a report containing the problem, affected element, reproduction steps, practical impact, repair and retest result. A score without that detail is hard to turn into a work plan.
Keep the result current
Rerun appropriate checks after changing templates, forms or interactive components. Include manual task checks when behavior changes. Save evidence for completed repairs and keep unresolved issues visible to the owner.
Our accessibility service is scoped around repairs and retesting. The guide to acting on an accessibility report explains how to prioritize findings. Neither a scan nor a limited repair engagement should be presented as proof that every part of a site meets every accessibility requirement.