Test your own site before anyone tests it for you

1. Unplug the mouse

Put the mouse away and complete one real task with the keyboard alone. Not the homepage: the thing people come for, end to end. Tab moves forward, Shift plus Tab moves back, Enter and Space activate.

Three failures show up immediately and all three are blocking:

• Focus you cannot see. If you lose track of where you are, so does every keyboard user. A visible focus indicator is not optional styling.
• A trap. Focus enters a widget, a modal or an embedded frame and cannot leave. Usually a video player or a chat box.
• Focus hidden behind a sticky header. This isWCAG 2.2 criterion 2.4.11, and sticky headers fail it constantly.

2. Listen to one page

Turn on the screen reader your machine already has and go through the page. Ourscreen reader guide covers the twenty-minute version, including how to pull up the heading list, which is the fastest way to see whether your page has a structure or only a layout.

3. Zoom to 200 percent

Set the browser zoom to 200 percent and then narrow the window to 320 CSS pixels wide. Text must reflow. Nothing may require horizontal scrolling, nothing may be cut off, and no control may become unreachable. Then set the browser's minimum font size up and check that your layout survives text scaling on its own, not only page zoom.

4. Check the forms, not the form fields

Every input needs a label that is programmatically associated with it, not a placeholder and not a paragraph sitting above it. Then trigger an error deliberately. The error has to be announced, it has to say what is wrong in words rather than colour, and focus has to move somewhere useful.

While you are there: can you paste into the password field? Blocking paste breaks password managers and fails WCAG 2.2 criterion 3.3.8.

5. Read the alternative text out loud

Not whether it exists – whether it says anything. Read each one aloud in place of the image and ask whether the sentence still works. Decorative images take an empty alt attribute so the screen reader skips them; an image that carries information needs a description of the information, not of the picture.

6. Run one automated scan, last

Now, and only now, run a free automated checker. Doing it last means you are not tempted to treat the score as the result. It will catch contrast values, missing language attributes and broken ARIA references quickly, which is genuinely useful and genuinely partial.

What to do with the findings

Sort them by whether they block a task, not by how many instances the tool counted.

1. Anything that makes a task impossible: unlabelled controls, keyboard traps, unreachable steps.
2. Anything that makes a task much harder: missing structure, poor error handling, focus problems.
3. Everything else: contrast, spacing, target size, alternative text quality.

Then write it down. For a public sector body that list is the basis of youraccessibility statement; for a private service under theEuropean Accessibility Act it is the evidence that you assessed the requirements rather than assumed them.

Where self-testing stops

This pass finds structural barriers. It does not tell you whether your service is usable by someone who navigates at three hundred words per minute with the screen off, or by someone using a communication aid, or by someone who needs the content ineasy-to-read. For anything where the stakes are real, one session with actual users is worth more than another round of tooling.