Most scanners read your page the moment it loads and hand you a list of CSS selectors. Your visitors don't use a page that way.
We crawl your whole site in a real browser, then do what a checklist can't: press Tab through every page, open the menus and dropdowns, shrink it to phone width, and measure colour contrast from the rendered pixels. Every issue is drawn on a screenshot of the page, so the person fixing it sees exactly what to change.
Standard: WCAG 2.1 Level AA (Title II)
EVERY ITEM LINKS TO A MARKED-UP SCREENSHOT
Illustrative report layout. Your numbers come from your site — and a score is never presented as a conformance certificate.
// why most reports don't get fixed
Automated accessibility reports tend to fail the same way: they are big, vague, and blind to the parts of a site people actually use.
Counts that hide the work. A report says 304 contrast problems. It is ten elements, six of them in the shared header — one edit fixes every page. Counting occurrences measures how far a problem has spread, not how much work it is.
A score measured against the wrong rule. Title II names WCAG 2.1 Level AA. Plenty of tools blend 2.2, AAA and "best practice" into one headline number, which then moves for reasons that have nothing to do with your legal exposure.
The page as it loads, and nothing else. Menus are closed. Nobody presses Tab. Nobody looks at the site on a phone. The dropdowns where residents pay a bill or find a permit go almost completely untested.
“Needs review: 3,000 items” is not a finding. It is the tool handing its uncertainty to your web team.
So we settle what can be measured, group what repeats, and send only the genuinely human judgements to a person.
// scored against the rule that applies
The DOJ's Title II rule requires WCAG 2.1 Level AA for state and local government web content. That is what the headline score measures — and it says so next to the number.
Title II score, WCAG 2.1 A + AA. Only failures of those criteria count. Newer WCAG 2.2 criteria, such as 24-pixel target size, are still found and shown — labelled as beyond Title II — and switch into the score with one setting if your policy adopts 2.2.
Level A and Level AA reported separately, as the share of pages with no automated failure at that level, with the change since your last scan.
Best practice beside the score, not inside it. Real defects that map to no WCAG criterion are listed, and never counted against conformance.
A site with a known failure never shows 100. An average can round a real failure away. Ours is capped so it can't.
An empty page is never a clean page. If a page fails to render — a firewall challenge, a slow script — it is marked as unjudged and left out of the average, instead of scoring a perfect 100 for having no content.
Overlays are reported, not credited. If your site loads an accessibility widget, we say so. Overlays are not accepted as Title II conformance, and a report that quietly scores around one is misleading.
// what we actually do to every page
Every page is rendered in a real browser — including navigation your CMS builds with JavaScript — and checked with the axe-core rule engine. Then come the checks a rule engine can't do on its own.
Check 01 · WCAG 2.1.1, 2.1.2, 2.4.7
Through the whole page, the way a keyboard user does. We flag focus you can't see, focus that lands on something hidden — the classic closed mobile menu whose links are still tabbable — keyboard traps, and controls that only respond to a mouse.
YOU SEE → The tab order, numbered on the page
Check 02 · any rule
Dropdowns, accordions, disclosure menus and lightboxes are clicked open and tested in their open state. Only what's new is reported, with a screenshot of the menu open. Links that navigate and submit buttons are never touched.
YOU SEE → The open menu, with the issue marked
Check 03 · WCAG 1.4.10
Each page is rendered at 320 pixels wide — the width WCAG uses for reflow. We catch sideways scrolling and name the element pushing past the edge, and flag text that gets cut off.
YOU SEE → A phone-width screenshot of every page
Check 04 · WCAG 1.4.3
When text sits on a gradient or a photo, rule engines give up and say "needs review". We hide the text, photograph what's behind it, and compare the text colour against every pixel. Most of that pile settles on its own — with the measured ratio shown.
YOU SEE → The ratio, and the colours measured
Check 05 · WCAG 1.1.1, 2.4.4, 1.2.2
An AI reviewer looks at each image beside its alt text — "DE4A1819" passes a rule engine and helps nobody — and suggests better text. We also flag vague links ("Read more") and videos to check for captions.
YOU SEE → A suggestion for a person to accept
The principle
Anything we can measure, we measure and show our working. Anything that needs judgement — whether alt text is right, whether a truncated heading loses meaning — goes to a review queue. AI suggestions are never counted in the score.
RESULT → A short, real to-do list
// the issue inspector
A web editor shouldn't need to read CSS selectors. Open any page and see its screenshot with each problem outlined and numbered, next to a plain explanation of what's wrong.
Numbered boxes on the screenshot match the numbered list. Click either one and the other follows.
Issues found at phone width or inside a dropdown are shown on that picture, not squeezed onto the wrong one.
Turn on the tab order and see where focus goes, with the stops where focus disappears marked in red.
One click re-checks a single page, usually in under a minute. What no longer reproduces moves to "Fixed" — no waiting for the next full crawl.
// built for fixing, not for counting
The report is organised the way a web team actually works through it.
Findings are grouped by the element that causes them. An element on every page is flagged "every page — one fix", because it lives in a shared header, footer or template.
Mark an item as not an issue, intended behaviour, or accepted risk — on one page or everywhere it appears. The decision, who made it and why carry into every future scan.
The WCAG criteria software can't decide are raised as a checklist per page template, so reviewing one article covers every article built the same way.
Score and Level A / AA trends over time, what's new and what's fixed since the last scan, pass rate per rule, and issues by click depth from your home page.
Weekly or monthly, automatically. New content is checked without anyone remembering to press a button.
Each WCAG 2.1 AA criterion with a combined verdict from the automated checks and your reviewers — supports, partially supports, does not support, or not yet evaluated.
// the rest of the site
Title II covers the documents you publish, not just the pages. The same crawl finds them.
Every linked document, found. PDFs and Office files are collected as the crawler goes, with the page that links to each one. Send them straight into bulk remediation, where every repaired file is independently validated against PDF/UA-1.
And why they're broken. Each document is traced to one root cause — the scanner's OCR setting, printing to PDF, the export preset, or how it was written — so you know whether the fix is a setting, a habit, or training, and which department it belongs to.
Broken links, checked once per scan. A link that goes nowhere is a dead end for everyone, and it shows up in the same report.
Your site, your permission. Scanning starts after you verify you own the domain, respects robots.txt, and identifies itself honestly. It is a guest on your web server, and it behaves like one.
One crawl: pages, menus, phone width, keyboard, and the documents behind the links.
// the honest limit
Automated testing, even ours, cannot certify a website as conformant. A vendor that implies otherwise is selling a number.
A machine can confirm an image has alt text, not that the text is right. It can confirm a form field has a label, not that the label matches the box. It can measure contrast, but not whether your reading order makes sense to someone listening to it.
So the score is described for what it is — no WCAG criterion failed an automated check — and the rest goes to people: the template review queue, the AI suggestions awaiting confirmation, and the items flagged "needs review".
A good scanner shrinks the human work to the part that genuinely needs a human. It doesn't pretend that part isn't there.
// how to compare scanners
Including us. None of these show up in a feature checklist, and all of them decide whether the report gets acted on.
| The question | Common answer | SentraCheck |
|---|---|---|
| Which standard is the headline score? | A blend of WCAG 2.0, 2.1, 2.2, AAA and best practice. | WCAG 2.1 A + AA, the Title II standard, stated beside the number. 2.2 is one setting away. |
| Are menus and dropdowns tested open? | Only as the page first loads, with everything closed. | Collapsed controls are opened and tested, with a screenshot of the open state. |
| Is keyboard access tested? | Only what can be inferred from the HTML. | Every page is tabbed through; the tab order is shown on the page. |
| What happens to "needs review" contrast? | Left for a person — often thousands of items. | Measured from the rendered pixels; most settle automatically, with the ratio shown. |
| Can I see where the issue is? | A CSS selector and an HTML snippet. | A marked-up screenshot, on desktop, phone, or with the menu open. |
| Do false positives come back every scan? | Often — dismissals are tied to one scan's results. | Decisions are kept per site and applied to every future scan, with who and why. |
// questions
No, and we will never present it that way. The score means no WCAG 2.1 A or AA criterion failed an automated check. Some criteria can only be judged by a person, which is why the report includes a review checklist per template and a Title II report that shows which criteria are still unevaluated.
You set the page limit per site, up to 50,000 pages. The crawler starts from your sitemap and follows links from there, so pages your navigation reaches are found even if the sitemap is out of date.
A scan loads a handful of pages at a time, like a few visitors browsing, and it respects robots.txt. You control how many pages it loads at once in the site settings.
In one place: reviewing whether alt text describes its image, and suggesting better text. Those suggestions go to a person to accept or dismiss, and they never count against the score. Every other check is deterministic and shows its working. See our security page for how data is handled.
We detect common overlays and say so in the report, but we don't credit them. The findings describe the underlying site, which is what the overlay is covering for, and overlays are not accepted as a substitute for conformance.
Yes. Re-scan a single page from the issue inspector. It usually takes under a minute, and anything that no longer reproduces is marked fixed.
Point us at your domain. You'll get every page scored against the Title II standard, every issue drawn on a screenshot, and a to-do list short enough to finish.