Why does automated accessibility testing not find every problem?

Because many WCAG requirements need judgement. A tool can see that an image has no alt text, but not whether the alt text it has is accurate; it can see a missing label, but not a confusing one. axe-core's makers say it finds on average 57% of WCAG issues. Use automated checks often and add manual testing.

Why automated accessibility testing finds some problems but not all

Published 15 October 2026 · General information, not legal advice

An automated accessibility checker reads a page's code and applies rules that have a clear yes or no answer. Many WCAG requirements do not. Whether a description is accurate, whether a link makes sense where it sits, or whether an error message helps someone fix their mistake are questions about meaning, and a person has to answer them. That is why every serious tool, Jralo included, says the same thing: automated testing finds many accessibility problems, not all of them.

What the published figures say

axe-core, the open-source engine behind many checkers, says it can find on average 57% of WCAG issues automatically.¹ When it cannot be sure, it reports the result as "incomplete" so a person can review it.¹
The UK Government Digital Service built a test page with 143 failures and ran 10 tools over it in 2017. All the tools together found 71% of the barriers; the best single tool found 37% from its errors and warnings, or 41% counting its prompts for manual inspection. GDS concluded that automated tools cannot be relied on by themselves to check a website.² Tools have improved since 2017, so treat the numbers as a picture of the limits, not as today's scores.
WebAIM found detectable WCAG failures on 95.9% of the top million home pages in 2026, an average of 56.1 per page. It adds that because only automatically detectable failures were counted, the share of pages that fully conform is likely lower still, and that a page with no detected errors is not necessarily accessible.³
The US Department of Justice says automated checkers can help but a clean report does not mean everything is accessible, and manual checks should accompany them.⁴

What a tool can decide, and what it cannot

Topic	A tool can usually tell	A person has to judge
Images	An image has no alt attribute; a linked image has no name.	Whether the alt text is accurate and useful, or whether the image is really decorative.
Forms	A field has no label in the code.	Whether the label is clear, and whether error messages explain how to fix the problem.
Colour contrast	The ratio between text and a solid background colour.	Text over photos, gradients and video, and whether colour alone carries meaning.
Links and buttons	A link or button is empty.	Whether "Learn more" makes sense where it is, and whether two links with the same text go to the same place.
Keyboard	Some signs of trouble, such as clickable elements that cannot receive focus.	Whether every task can be finished with a keyboard, in a sensible order, without getting stuck.
Headings and structure	Headings that skip levels or are empty; a missing page language.	Whether the headings describe the content and the reading order makes sense.
Video	A video element has no caption track.	Whether captions are accurate and visual information is described.
Consistency and help	Very little.	Whether help appears in the same place on each page, and whether a process asks for the same information twice.

The pattern is the same throughout: a tool can confirm that something exists and has the right shape in the code. It cannot tell whether it means the right thing to a person. WCAG itself is written as outcomes for people, such as that a text alternative "serves the equivalent purpose", which is a judgement, not a measurement.⁵

Why automated checks are still worth running

They are fast and repeatable. A check that takes seconds can run every day or on every release, which no manual audit can match.
They catch regressions. Stores change constantly: new products, new apps, theme updates, seasonal banners. A new problem from any of them shows up on the next check.
They find the common problems. The most frequent failures WebAIM reports, such as low-contrast text, missing alt text, missing form labels and empty links, are exactly the kind tools detect well.³
They give you a record. Dated results show what was tested and when, which matters if you ever need to show your work.

A sensible way to combine them

When	Automated	Manual
Every day, or every release	Check the key pages: home, a category, a product, contact.	Not needed unless something changed.
After a new theme, app or redesign	Check again straight away.	Keyboard test and a short screen reader pass of the changed pages.
After any change to cart, checkout or sign-up	Where the tool can reach them.	Buy something with only a keyboard, then again with a screen reader.
Once or twice a year	Full site.	A review of content: alt text, link text, headings, video captions.

Jralo's checks skip cart, checkout, account and sign-in pages on purpose, so an automated run never adds to a cart or signs anyone out. Those pages are where a barrier costs you a sale, so they always need a manual test.

Reading an automated report well

Start with critical and serious issues on pages that lead to a purchase or an enquiry.
Look for patterns. Twenty identical errors on twenty product pages usually mean one problem in one template.
Review "needs review" or "incomplete" items yourself rather than ignoring them; they are the tool telling you it cannot decide.
Re-test after fixing and keep the result, so the record shows the problem going away.

What this means for claims

Because tools cannot find everything, no automated checker, score or badge can show that a site meets WCAG or any law. Be wary of any product that says otherwise. Jralo runs axe-core against WCAG 2.2 A and AA on your pages every day on paid plans and keeps a dated record of each check, and its reports say plainly that automated testing finds many, not all, problems. For the manual part, how to check your website in 15 minutes walks through a keyboard and screen reader pass, and how to write alt text that helps covers one of the judgement calls tools leave to you.

Questions people ask

If an automated check finds no problems, is my site accessible?

Not necessarily. WebAIM notes that the absence of detected errors does not mean a page is accessible or conformant, and the US Department of Justice says manual checks should accompany automated ones. A clean result means the problems a tool can detect were not found.

Do automated checkers give false positives?

They can. axe-core is designed to report zero false positives, with the caveat "bugs notwithstanding", and marks results it cannot decide as incomplete so a person can review them. Other tools make different trade-offs, so read a result before you change code.

How often should I run automated checks?

Whenever the site changes, and ideally every day for a store whose products, apps and content change often. Automated checks are cheap to repeat, which is their main strength. Manual testing can be less frequent: after a redesign, a new theme, or changes to checkout.

Check your site now

Jralo's free check tests your homepage and up to five pages for accessibility (WCAG 2.2 A and AA) and website QA problems, and tells you which countries' accessibility laws are likely relevant to you. No sign-up.

Run the free check

Sources

Every number and legal statement in this post links to one of these sources, read on 13 October 2026.