What automated accessibility testing checks, and what it cannot

The short answer: automated tests find many common accessibility problems quickly and the same way every time: missing image descriptions, unlabeled form fields, low color contrast, controls you cannot reach with a keyboard, and PDFs a screen reader cannot read. No automated test finds every problem, and we say so.

The standard

Both federal web rules, the Justice Department's ADA Title II rule and the HHS Section 504 rule, use WCAG 2.1 level AA as the technical standard. Our tests check web pages against it. For PDFs, we check whether they have tags, real text, a title and a language. Within your package we rebuild PDFs that lack tags or real text, and give instructions for adding a missing title or language.

What our automated tests check

  • The WCAG 2.1 A and AA rules that the axe-core testing engine checks, for example missing text descriptions on images, form fields without labels, color contrast, the page's language, heading and landmark structure, and names for buttons and links.
  • Keyboard-only use. The test presses Tab through each page the way a person who cannot use a mouse would, and reports three problems: something you can click that Tab never reaches; Tab getting stuck inside part of the page; and a place where Tab lands but nothing visibly changes. Controls inside frames from other websites are not tab-tested, and the report says so.
  • Descriptions and link wording. Image descriptions that exist but say nothing useful, such as "image", "photo" or a file name, and links whose words do not say where they go, such as "click here" or "read more".
  • Form errors (paid audits). On up to five forms per page, the test presses the submit button with the required fields left empty and checks that each field gets an error a screen reader can reach. It never types anything, and it blocks anything the form tries to send, so nothing reaches your site. Free scans skip this test.
  • PDFs. Whether each document has tags, real text or only page images, a title and a language.

What automated testing cannot judge

Software cannot fully judge whether a description is meaningful in context, whether captions and transcripts are accurate, whether instructions make sense, or how easy a page really is to use with a screen reader. Some limits are specific enough to name:

  • Captions and audio description. For video and audio played directly on the page, we check that a caption track exists, not whether the wording is accurate or complete. We do not check for audio description, or for captions on videos embedded from services such as YouTube or Vimeo.
  • Reflow at 400% zoom. We check for content that forces side-to-side scrolling at a 400%-zoom-equivalent width, not whether every feature still works and stays easy to use at that size.
  • Form error handling. We check that each empty required field gets an error a screen reader can reach and that fields are labeled, not whether the error message actually explains what to fix.
  • Meaningful sequence (reading order). We check the order elements sit in the code, not whether that order actually makes sense to someone using it.
  • Timing and flashing content. We check for content that flashes fast enough to be a seizure risk, not every case of timing that may be too short, such as a countdown or a session that expires quickly.

Automated tests find a large share of common problems, not every possible one.

What our crawler cannot reach

Our crawler has no way to sign in. It cannot see anything behind a username and password, a one-time code, or any other login: a patient portal, an online-scheduling account, a resident self-service account, or a staff-only page. It only tests pages and documents that anyone can already reach without signing in. If your most important content lives behind a login, tell us in your order and we will say plainly in your report what we could and could not see.

How to use the results

For each problem, the fix pack gives your web team corrected code where HTML and CSS can fix it, specific instructions where code alone is not enough (a keyboard trap in a script, for example), and suggested wording for image descriptions, labels and links, written by AI. When the AI is unsure, it gives an instruction instead of guessing. Once the fixes are live, re-test, and keep the written test report with your records. Our guarantee: if, once your web team has applied our fixes, the covered pages do not pass our full automated WCAG 2.1 AA tests by your deadline, we keep writing fixes and re-testing at no charge until they do.

EveryRamp tests websites and PDFs against WCAG 2.1 AA and writes the fixes. You can start with a free scan of your website, see a sample report, or compare the packages.

This page is general information, not legal advice.

Guides