Let us say you have just joined a team and your lead says, “Run axe on the checkout page and tell me if it is accessible.” You run it, it reports zero violations, and you feel pleased. Hold on to that feeling for a minute, because we are going to test it.

Accessibility testing checks whether people with disabilities can use your product: someone who is blind and uses a screen reader, someone who cannot use a mouse, someone with low vision or colour blindness, someone with a hearing impairment watching a video, someone with a tremor who struggles with tiny tap targets. The yardstick most teams use is WCAG, the Web Content Accessibility Guidelines from the W3C. They come in versions (2.1 and 2.2 are the ones you will meet most), and in three levels: A, AA and AAA. Level AA is what contracts, tenders and most policies ask for. In India, the Rights of Persons with Disabilities Act, 2016 is the legal backdrop, and government sites follow GIGW, the Guidelines for Indian Government Websites, which lean on WCAG.

What a scanner is good at

Scanners such as axe-core (the engine inside many tools, and behind the accessibility audit in Lighthouse), WAVE or Pa11y read the page’s DOM and apply rules. They are excellent at things that can be decided by looking at markup and computed styles:

  • An image with no alt attribute at all.
  • A form field with no programmatic label.
  • Text whose colour against its background falls below the AA contrast ratio (4.5:1 for normal text, 3:1 for large text).
  • Duplicate id values, missing page language, invalid ARIA attributes, buttons with no accessible name.

These checks are fast, cheap and repeatable, which makes them ideal for a CI pipeline. Run them on every key page on every build and you stop a large class of regressions from ever reaching a tester. That is real value, and nobody should sneer at it.

What it quietly cannot judge

Here is the uncomfortable part. A scanner can tell that an image has alt text. It cannot tell that the alt text says “IMG_4021.jpg” or “image” or, on a product page, the same sentence for all eight photos. The rule is satisfied; the user is not. Estimates of how much of WCAG can be checked automatically vary depending on who counts and how, but the honest summary is that it is a minority to roughly half of the issues, never all of them. A clean report means “no machine-detectable failures”, not “accessible”.

Things the tool cannot decide:

  • Whether the alt text, link text or heading actually describes its purpose. “Click here” passes many rules.
  • Whether the reading and focus order makes sense. A page can be perfectly valid and still jump from the header to the footer to the middle.
  • Whether a custom dropdown built from div elements can be opened with the keyboard and announces its state.
  • Whether an error message is announced when a form fails, or only appears in red.
  • Whether captions on a video are accurate, or whether an animation can be paused.
  • Whether instructions rely only on colour, shape or position (“press the green button on the right”).

The three manual checks that find the most

You do not need a specialist lab to start. Do these three on the same checkout page, and see how your clean report holds up.

Keyboard only. Put the mouse away. Use Tab, Shift+Tab, Enter, Space and the arrow keys to complete a purchase. Can you reach every control? Can you always see where focus is? Does a pop-up trap you, with no way to close it using Escape? Is the order of focus the order of the screen? This takes ten minutes and routinely finds blockers, because anyone who cannot use a mouse, and every screen reader user, depends on it.

A screen reader, even badly. NVDA is free on Windows; TalkBack ships on Android; VoiceOver is built into iPhones and Macs. You will be slow at first, and that is fine. Close your eyes or turn the screen off and listen. Does the heading structure tell you what is on the page? When you add a coupon, does anything announce that the total changed? Does the pin-code field say what it is, or only “edit text”?

The third check needs no tool: zoom the browser to 200 percent, then try a narrow phone width. Text should reflow without hidden content or sideways scrolling for every line. Add a larger system font on Android and see whether buttons clip their labels.

Back to the checkout page

Suppose your zero-violation page has a custom “Select delivery slot” widget. The scanner saw valid markup. But with the keyboard, the slots cannot be focused, because they are click handlers on plain elements. A screen reader user hears nothing at all where the widget sits. The cure is usually to use native controls, such as radio buttons or a real select, before reaching for ARIA. That is a general principle worth remembering: no ARIA is better than bad ARIA, and a native element gives you keyboard support and semantics for free.

Mixed teams get the best result by layering. Linting in the editor catches the easy things as code is written. Automated scans run in CI and block new failures. A short manual pass with keyboard and screen reader happens on each important new flow, and a deeper audit, ideally including people who use assistive technology daily, happens before big releases. Keep a record of which WCAG success criteria were checked by which method, so that “we tested accessibility” has a precise meaning.

Finally, be careful with the word “compliant”. Overlay widgets that promise one-line accessibility fixes are widely criticised by accessibility practitioners, because they cannot repair broken structure. Treat a scanner as a smoke detector: very good at telling you there is smoke, and no use at telling you the house is safe to live in.

Leave a Reply

Trending