Skip to content
A11ySignal
Sign inCheck a page

How it works

From one URL to an alert that is worth reading

No agent to install, no snippet to add to your site. You give it an address; it behaves like a careful visitor.

01

It works out which pages to check

The crawler starts at your URL, reads your sitemap if you publish one, and follows links on the same site from there. Sitemaps matter more than people expect: they surface pages that are only reachable from search or pagination, which is exactly where forgotten templates live.

readssitemap.xmlIncluding a sitemap index, one level deep.+ index
honoursrobots.txtDisallowed paths are never fetched. Crawl-delay is respected up to five seconds.+ crawl-delay
skipsassetsPDFs, images and archives are not pages, so they are not opened.by extension
dedupesurlsTracking parameters are stripped and query order normalised, so one page is scanned once rather than four times.normalised
02

Each page opens in a real browser

Pages load in headless Chromium at 1280×900 with JavaScript running. The scanner waits for the network to settle before it measures anything, so a client-rendered app is checked as a user would see it — not as an empty shell. Then axe-core 4.13.0 runs against every element in every frame.

03

Findings are compared with last time

This is the part that makes monitoring different from auditing. Every scan is diffed against the previous one for the same site: what is new, what got fixed, which pages appeared or disappeared, and how the score moved. A failure that was already there does not get reported again — you would learn to ignore the emails, and then you would ignore the one that mattered.

newregressionA rule that now fails, or fails on more elements than before.alerts
fixedprogressCounted and kept, so you can show a client the line moving.recorded
unchangedknownStill in the report. Never in the alert.quiet
04

You hear about it once

Email and Slack fire on the diff, not the total. The message leads with the worst new issue, names the page and the rule, and links to the report. If nothing changed, nothing is sent.

Where the limits are

Pages behind a login are not scanned yet. Automated rules cover roughly a third of WCAG — see how the score is calculated for exactly what that means. And a scan tells you what fails, not whether your alt text is any good.

Try it on one page