How it works
What actually happens when you enter an address, what the score means, and where the limits of automated testing are.
The test
Your page is loaded in a real headless Chromium browser at a 1366×900 viewport. We wait for JavaScript to execute and the network to settle, so single-page applications and client-rendered content are tested as a user would actually receive them — not as raw source HTML.
We then run axe-core (opens in a new tab), the open-source rule engine maintained by Deque and used inside Chrome DevTools, Lighthouse and most commercial accessibility platforms. We run the rules tagged for WCAG 2.1 Levels A and AA — the conformance target named by EN 301 549, the harmonised European standard the EAA points to.
Rules tagged best-practice are collected but reported separately as advisory, because they are not conformance failures and we are not going to inflate your issue count with them. WCAG 2.2 criteria are also collected separately as forward-looking.
The crawl
A free scan tests one page. A paid audit crawls the site, seeded from your sitemap where one exists and otherwise by following links from the homepage.
Candidate pages are ranked before testing, so a limited page budget is spent on the journeys that matter: checkout and cart, contact, signup and login, pricing and product pages. Blog archives are ranked last, because two hundred posts sharing one template contribute one template's worth of information.
We respect robots.txt, honour Crawl-delay, and pause between requests.
More about the crawler.
The score
The 0–100 score weights each finding by severity and by how many elements it affects, then normalises per page so a five-page site and a fifty-page site are comparable.
Repeated instances of the same rule count on a curve rather than linearly. Forty elements failing contrast because of one CSS variable is one fix, not forty problems, and scoring it linearly would make every large site look identically catastrophic — which would make the number useless.
Be clear about what this number is: our metric, for tracking one site over time. It is not defined by WCAG, EN 301 549 or any regulator, and it should never be quoted as a conformance level. The conformance verdict on the report is the part that maps to the standard.
What this cannot tell you
Automated testing reliably covers somewhere around a third of WCAG success criteria. It is excellent at the objective, machine-checkable ones: contrast ratios, missing labels, missing alt attributes, invalid ARIA, document structure, language declaration.
It cannot tell you whether:
- your alt text is meaningful, rather than merely present
- your focus order makes sense to a person
- your error messages are actually helpful
- your custom date picker, modal or combobox works with a screen reader
- a blind user can complete a purchase from start to finish
That last one is where most real risk sits, and no scanner will ever answer it. A clean automated result means the machine-detectable failures are gone — a necessary step, not a finish line. Anyone selling you a script as a compliance certificate is selling you a liability.
What we store
Scan results, the URLs tested, and the markup snippets of failing elements. No cookies from the scanned site, no form submissions, no content behind a login — we test what an anonymous visitor sees. Reports are unlisted and excluded from search engines. Full privacy detail.