Automated accessibility testing
One crawl of your site.
One list of what to fix first.
Point it at your homepage. It follows your sitemap and your links, tests every page it reaches, then groups thousands of raw findings into a short, ordered list of the actual changes to make — each one with the code that fixes it and the pages it clears.
Test five pages free. No account, no card.
Enough pages to show which defects repeat across your
templates, in well under a minute. We obey robots.txt and identify ourselves
as SitePunchlistBot.
Buttons with no readable text
The fix
<button aria-label="Close dialog">
<span class="icon-close" aria-hidden="true"></span>
</button>
8 of those pages share one component.
The same element fails in the same place on each, so a single change clears
all 8..icon-btn:nth-child(n)
Read the whole sample report — 320 defects across 13 pages.
The problem with a raw scan
A scanner on a 400-page site returns something like 9,000 findings. That is not a plan of work — it is a reason to close the tab.
Findings are not tasks
The same broken navigation counts once per page. Four hundred “findings” are one component and one afternoon. We detect when a defect shares the same position across pages and say so: one fix, 400 pages.
Severity is not priority
A critical defect on a page nobody visits matters less than a serious one in the header of every page. We order by severity and reach, and label who fixes each one — developer, content editor, or designer.
What a paid crawl returns
- A prioritised fix list. Ordered by severity and how many pages each defect touches, not by rule name.
- The code change for each defect. A before and after, in plain HTML or CSS, written to be pasted into a ticket.
- Every affected page and selector. The exact CSS path to each failing element, so nobody has to hunt.
- A dated testing record. What was tested, when, by which engine version, against which rule set — and which pages could not be reached.
- A shareable report and a PDF. Plus CSV, for anyone who wants the raw rows in a tracker.
Coverage is stated, not implied
Every report says how many pages were discovered and how many were tested. If your site has 4,000 pages and your plan covered 150, the report says so on the first page.
It also lists the pages that failed to load, and any the site's firewall refused — silence about those would make the total meaningless.
What we cannot test
Every vendor in this market has an incentive to be vague about this. Here is the number, computed from the engine we actually run.
Automated rules catch a large share of the raw volume of defects — repeated missing labels, contrast, unnamed controls — because those are the ones that repeat. They catch a minority of the criteria. Keyboard traps, focus order, whether alt text is any good, whether an error message helps: a machine cannot judge any of it.
See the full table — all 55 criteria, and which of them we can reach.
Pricing
One payment. No subscription, no widget, nothing to install on your site.
Site Punchlist
$49 one payment
- Up to 150 pages
- Full prioritised fix list
- Hosted report, PDF and CSV
- Ready on your order page when it finishes
Site Punchlist + Rechecks
$149 one payment
- Up to 750 pages
- Everything in Site Punchlist
- A fresh crawl each month for three months
- Each recheck shows what was fixed, what is new, what remains
What this does not tell you
Automated testing cannot establish WCAG conformance, and it cannot establish legal compliance. Nothing here is a certification or a legal opinion.
The rules we run reach 23 of the 55 WCAG 2.2 Level A and AA success criteria. 32 have no automated rule at all — keyboard traps, focus order, focus visibility, error identification and consistent navigation among them. Those need a person. See the full table of what we can and cannot check.
Questions
Does this make my site compliant?
No, and no automated tool can. Compliance depends on the whole site, on testing a machine cannot do, and on rules that differ by country and by the kind of organisation you are. What we sell is findings and a dated record of what was tested. What you do with that is the work.
Is this an overlay or a widget?
No. Nothing is added to your site and nothing runs on it. We load your pages the way a browser does, run the rules, and leave. The fixes go in your code, where they last.
How long does a crawl take?
Between two and five seconds a page, depending how heavy your pages are — so usually five to fifteen minutes for 150 pages, and up to about an hour for 750. You get an email with the link when it is done; there is nothing to sit and watch.
What if the crawl fails or finds nothing?
If we cannot reach enough of your site to produce a useful report — a firewall blocks automated testing, or the address does not resolve — we refund you without argument. The refund terms are here.
Who is behind this?
A one-person business. No sales team, no call to book. Buy it, and the report arrives on its own.