The QA Accessibility Test Pass

Run a scanner on every page in the journey.

Scan each state as well as each page, which means opening the dialogs, submitting the form wrong, and scanning the error state too. A scanner sees one state of one page and never reaches the others on its own, so the states you skip are states nobody has tested.

Treat every scanner finding as a lead rather than a verdict.

Confirm each one by hand before you file it, because a defect a developer disproves in five minutes costs you the credibility of the next ten. Contrast over an image is where false positives cluster. The tool is guessing at a background that varies pixel by pixel.

Write down what the scanner could not see.

Meaning, reading order, and whether alt text is true are all invisible to automation, and so is anything that needs a second state of the page. A clean scan is not a pass. Saying so in the test notes stops somebody quoting it as one in a release meeting.

Complete the whole journey with the keyboard.

Start to finish, including submitting. Anywhere you get stuck is a defect with a clear reproduction.

WCAG 2.1.1

Watch the focus indicator the entire time.

Note every point where it disappears, becomes invisible against the background, or jumps somewhere unexpected.

WCAG 2.4.7

Open and close every overlay.

Dialogs, menus, date pickers and tooltips, and each has its own expected behavior. A modal should close on Escape and usually hand focus back to what opened it, unless the workflow has moved on. A tooltip normally leaves focus where it was and has to be dismissable without moving the pointer. What the rule requires everywhere is that you can get out with the keyboard, not that Escape is the key.

Count the Tab presses to the main content.

Load a fresh page and count. A handful of presses tells you something is already carrying people past the navigation. Forty tells you to go looking for the mechanism, not that there is not one, because a landmark route works without costing a single press. Neither number is a threshold in the rule. Do not run the version of this test that says press Tab once and look for a skip link, because it reports a failure on a conforming page. Landmarks and a heading on the main content each satisfy this rule on their own, and nothing requires a skip link to be the first focusable thing on the page.

WCAG 2.4.1

Listen to every form field.

Each should announce its label, its type, and whether it is required. 'Edit text, blank' is a defect.

WCAG 4.1.2

Submit the form wrong and listen.

If the errors are silent, that is a defect even though they are visible on screen.

WCAG 4.1.3

Listen to every button and link.

Anything announced as just 'button' or 'link' has no accessible name.

WCAG 4.1.2

Pull up the heading list.

It should read as an outline of the page. Anything that looks like a heading on screen and is missing from the list is a 1.3.1 defect. A page with genuinely no headings is a usability problem rather than a rule breach at AA.

WCAG 1.3.1

Set the viewport to 320 pixels.

That is the width the reflow rule is measured at, and it is also a real phone. Look for horizontal scrolling, for content clipped at the edge of its container, and for controls that have slid off screen entirely. Anything you have to scroll sideways to read is the failure this rule exists to catch.

WCAG 1.4.10

Zoom text to 200 percent.

Text-only zoom is the sharper test, because page zoom scales the layout along with the words and hides the problem you are looking for. It is not the only conforming route, though, so do not file a failure against a page that meets this through full-page zoom. What you want is overlap, and text cut off inside a container with a fixed height, which is where this fails almost every time.

WCAG 1.4.4

Check contrast on every state.

Default, hover, focus and error, because a component is only as accessible as its worst state and the default is the one everybody checks. Placeholder text is the usual failure. Disabled controls are excepted by the rule itself, so a grayed-out button measuring short is not a defect to file.

WCAG 1.4.3

View the journey in grayscale.

Drop the color out and walk the journey again, watching for anything that has just become ambiguous. Status pills, required-field markers, validation states and chart series are the four that usually go, and each one needs a second signal rather than a different color.

WCAG 1.4.1

Turn on reduced motion and repeat.

Switch the operating system setting on, reload, and walk the journey once more. Large movement should be cut back or gone. This is the check skipped most often, and for somebody with a vestibular disorder the motion it catches causes real nausea rather than mild irritation.

Name the barrier rather than the rule.

"A keyboard user cannot reach the close button on the delivery dialog" lands with a developer. A bare criterion number sends them to look something up before they know whether it matters. Put the barrier in the title and the criterion in the body.

Give the exact steps and the assistive technology used.

Browser, screen reader and both version numbers, because behavior differs between them and a developer who cannot match your setup cannot see what you saw. This one line is the difference between a defect that gets fixed and one that gets marked as not reproducible.

Cite the criterion so severity is not a debate.

Link the rule page in the ticket. It turns an argument about whether something is worth doing into a question with a documented answer, and it gives whoever triages the ticket something they can point at when they schedule it.

Say who it affects.

Name the group and name what happens to them. Triage is honest once that line is there, because a blocker for screen reader users stops reading as a cosmetic issue the moment somebody has to write down that it stops those customers buying.