Exit course
Chapter 10 of 10 Intermediate 7 min read Updated

Knowledge check

Nine questions on the decisions that come up in real accessibility testing work. Each answer has a short explanation, so a wrong answer points you back to the right chapter.

1. A product owner asks you to "make sure the site meets WCAG 2.2 AA with automated tests". What is the accurate response?

Rules engines are good at mechanical criteria such as contrast, names and labels, but many criteria need human judgement: alt text quality, focus order, meaningful link text. Automation and manual checks complement each other (chapters 1 and 8).

2. A scan written two years ago uses withTags(['wcag2a', 'wcag2aa']). What does it miss?

Each WCAG version's criteria have their own tags. Without wcag21a, wcag21aa, wcag22a and wcag22aa, rules like target-size (2.5.8, new in 2.2) never run (chapters 2 and 3).

3. A test opens a settings dialog, then immediately calls analyze(). The dialog's contrast results vary between runs. What is the best fix?

axe measures what is rendered when it runs. A dialog still animating in has partial opacity, which lowers contrast. Scan the settled state; retries and disabled rules only hide the timing problem (chapter 6).

4. An embedded third-party map widget fails aria-allowed-attr, and your team cannot change its markup. The vendor has a ticket. What is the most defensible setup?

A scoped, documented exclusion keeps the rule active for everything you own and makes the vendor issue visible. Disabling the rule or a whole tag would also hide your own future violations (chapter 4).

5. A pull request adds three lines to a11y-baseline/checkout.json and nothing else changes in the checkout code. What should the reviewer do?

A baseline entry is permission for a violation to exist. New entries are either newly introduced bugs or selector churn after a markup change; both deserve a look before they are accepted (chapter 4).

6. A custom toggle is <span role="switch" tabindex="0"> and axe reports aria-required-attr. What is the most robust fix?

A static aria-checked silences axe but lies about state as soon as the user toggles it. Native inputs bring focus, Space handling and real checked state; removing the role or hiding the control makes it worse (chapter 5).

7. Your suite passes in Chromium. On a Safari run, a button fails target-size. What is the right reading of that result?

axe measures the layout each engine produces. Different default control sizes are a real difference for real users, which is one reason to run accessibility scans in more than one browser (chapters 5 and 7).

8. Which problem is a scheduled scan of production URLs most likely to catch that pull-request tests miss?

Scheduled scans re-check pages that change without a code change, such as CMS content and third-party scripts. The other three need human judgement, which no scan provides (chapters 7 and 8).

9. A designer swaps a visible "Email" label for a placeholder. axe stays green. Which test would catch the change?

axe accepts a placeholder as an accessible name, so the rules stay green. A snapshot of the accessibility tree, or a test that checks the visible label, encodes the decision that the field must keep a real label (chapter 8).

Where to go next#

Put the four layers from chapter 8 in place on one real page of your own product this week: an axe scan with a baseline, one keyboard journey, one accessibility tree snapshot of its most important form, and a fifteen-minute manual pass with a screen reader. Then grow from there.

For reference, keep the W3C WCAG 2.2 quick reference and TestingBot's accessibility rules overview open while you fix issues. To run your checks across more browsers, or to scan production pages on a schedule, see automated accessibility testing on TestingBot. If you want to know more about the test runner used throughout this course, the Playwright tutorial goes deeper.

You finished Accessibility Testing Tutorial: WCAG 2.2

Put what you learned into practice. Run your tests on 6100+ real browsers and physical iOS and Android devices.