Accessibility measured on every crawl, not once a year
An audit delivered as a PDF describes a site that no longer exists. On every scheduled run we replay the checks a machine can make, show the failing element as an image, and suggest the fix whenever it can be computed, so that compliance becomes a tracked state rather than a one-off project.
Access opens in waves: we email you when yours is ready.
Our numbers, with their provenance
-
WCAG 2.2 AAreference
-
RGAA 4.1reference
Compliance is dated the day someone looked
Between two audits a live site changes by several hundred pages. The signed document then describes a state nobody visits any more, and its evidentiary value erodes with nothing to signal it.
-
01
The annual audit expires at the first deployment
The report lands, fixes get scheduled, and three hundred pages ship in the meantime. Nobody knows whether they meet what the report required.
-
02
Findings do not say where to look
A criterion identifier and a URL do not help a developer find the failing element inside a two-thousand-node page. The ticket gets opened, then it waits.
-
03
The deadline arrives before the tooling
The regulatory obligation carries a date; the ability to demonstrate where you stand takes several quarters to build. Many teams discover the gap far too late.
What we measure
What the audit records
Checks run against the page as a browser builds it: on the rendered document rather than the initial HTML. On a modern application that difference decides half the findings.
- Automatable violations (the analysis engine runs on every page of the crawl. Each finding carries its criterion, its level and the exact element concerned, down to the offending node.)
- Contrast cards (every insufficient text/background pair is measured on the actually composited colour rather than the declared value. A fix is proposed when one exists: the nearest shade that clears the threshold.)
- A screenshot of the failing element (an image of the element in context accompanies the finding. That is what turns a criterion identifier into a task a developer can pick up without an investigation.)
- The compliance report (an exportable document tied to the reference versions our registry publishes, which also names the checks that could not be run.)
Further reading
EAA, WCAG, VPAT: what accessibility compliance actually asks of a web team
The guide: what compliance actually requires, across EAA, RGAA and VPAT.
Read ComparisonNessFlow vs Screaming Frog
What a technical crawler covers, and what it does not say about accessibility.
Read GlossaryGlossary
RGAA, EAA, VPAT, contrast ratio: the definitions, without shortcuts.
ReadOther modules
Frequently asked questions
How much of the criteria can a machine judge?
A minority, and that is true of every automated engine, not just ours. An automated check detects a missing attribute, an insufficient contrast, an inconsistent heading structure. It cannot tell whether alt text really describes the image, or whether a keyboard journey stays usable to the end. So we display the checks that ran and name what was not judged.
What does the report contain?
Findings by criterion and by page, with their level, the element concerned and its screenshot; the list of checks that ran; and what requires human evaluation. It is exportable and tied to the reference versions our claims registry publishes: if we move to a new version, the document says so, because a report dated against a version it does not name cannot be relied on.
Does the audit run on authenticated pages?
Not by default: the crawl covers what is public. A protected area requires explicit configuration on your side, and we would rather it stayed a deliberate decision than a quiet setting nobody remembers six months later.
Does a high score mean I am compliant?
No, and this is the most important question on the page. The score covers what was measured. An axis that could not be evaluated is removed from the denominator and flagged as not measured: it is never counted as passing, because a renormalisation that erased a known defect would raise the grade of a site whose problem is still open. Conformance is declared, with a share of human evaluation we do not replace.
The exact scope, and this is the important part
A machine can only judge some of the criteria. A page with no automatable violation is not an accessible page, and our report says so in plain words rather than letting a high score imply the opposite.
What we cover
- Machine-judgeable criteria, replayed on every scheduled crawl
- Contrast, measured on the composited colour rather than the declared value
- An exportable report that names the checks actually run
Stated limits
- No statement of conformance: a report records findings, it does not certify
- Whether alt text is meaningful, reading order, keyboard journeys: human judgement
- No changes applied to your code, a suggestion stays a suggestion
- No testing with real users, which remains the only proof of use
Measure the real state before committing to a date
A first run gives you the inventory of automatable violations on your public pages, with screenshots. That is the starting point of a plan, not a certificate.
Access opens in waves: we email you when yours is ready.