EN 301 549 · WCAG 2.1 AA · European Accessibility Act
Accessibility findings your developers can actually fix.
We scan your client sites on a weekly schedule, then hand back a runnable Playwright suite where each defect is a failing test: red now, green only when it is genuinely fixed. Not a PDF. Not an overlay widget.
Send one URL you are authorised to test. We reply with the real findings and the exact fixes. No account, no card.
Why this is urgent now
The deadline has already passed.
The European Accessibility Act became enforceable on 28 June 2025. It binds a defined list of sectors (e-commerce, banking and payments, ticketing, passenger transport, telephony and audiovisual media) across all 27 member states, to EN 301 549, which takes WCAG 2.1 Level AA as its web baseline. The first French lawsuits were filed in November 2025.
| Member state | Exposure |
|---|---|
| Sweden | up to approximately €900,000 |
| France | €5,000 to €250,000, plus €25,000 per year for a missing accessibility statement |
| Germany | up to €100,000 per violation |
| Ireland | up to €60,000 |
The Act also requires a published accessibility statement, an accessible support channel, and a documented, ongoing compliance process. Conformance drifts with every deploy, so a single audit satisfies none of it.
It does not apply to everyone, and we will tell you if it does not apply to you. The Act defines e-commerce services as those provided at the request of a consumer with a view to a consumer contract, so business-only services fall outside it. Article 4(5) separately exempts microenterprises providing services, meaning fewer than 10 people with turnover or balance sheet under 2 million euro. If your client is exempt, the legal argument for this work does not hold, and the remaining reasons are the ones you would give anyway: customers who cannot check out, and procurement questionnaires that ask.
Penalty figures above are maximum exposures set by national implementing legislation and are reproduced from published summaries, not from primary legislation we have read. Treat them as indicative. Your client's exposure depends on their member state, sector and size.
How it works
Four artifacts, generated on every run.
-
You connect a site
A URL you are authorised to test. We crawl its public pages: read-only, respecting robots.txt, no authenticated areas, no form submissions, no load testing.
-
We run the checks
axe-core against the WCAG 2.1 A and AA rulesets in both light and dark colour schemes, plus functional checks: JavaScript console errors and failed requests. Every failure keeps the exact DOM node, CSS selector, a screenshot and a Playwright trace.
-
You get code, not a document
A git repository wired for CI where every finding is one failing assertion. Your developers run it, fix, and watch it go green. Remediation is verified by the build.
-
Artifact 01
Playwright suite
A real repo with a GitHub Actions workflow. One assertion per finding, traces and screenshots retained on failure.
-
Artifact 02
Computed remediation
For contrast failures, the replacement value is solved against the measured background:
#61b5aaat 2.27:1 becomes#427b74at 4.58:1, not "increase contrast". -
Artifact 03
White-label report
Client-ready under your agency's branding, so you forward it rather than rewrite it.
-
Artifact 04
Evidence bundle and history
Screenshots, Playwright traces and the raw scan, plus a dated log per site showing whether the defect count moved since the last run.
Our own measurements
We scanned 21 EU shops. Thirteen of the fourteen we could measure properly had defects.
Targets were drawn from the public portfolio pages of German, Swiss, Dutch and French web agencies, which is work those agencies chose to showcase. Scans were capped at three pages per site to keep load on third-party servers light, so every figure here is a floor. Seven sites returned one page or fewer and are excluded rather than counted as clean: a blocked crawl and a conformant site produce identical zeros.
We publish aggregates only. We never name a site's defects publicly, and we never approach anyone by threatening to. Findings go to the site's owner or to whoever commissioned the scan.
| Rule | Criterion | Sites | Elements | Severity |
|---|---|---|---|---|
color-contrast | 1.4.3 | 10 | 93 | serious |
image-alt | 1.1.1 | 10 | 67 | critical |
button-name | 4.1.2 | 7 | 39 | critical |
link-name | 4.1.2 | 8 | 34 | critical |
select-name | 4.1.2 | 5 | 30 | critical |
aria-required-children | 1.3.1 | 6 | 14 | critical |
select-name and button-name are the ones that matter commercially.
On a shop, those are variant pickers and add-to-cart controls, which is the checkout path. An
unlabelled select is not a styling nitpick; it means a screen-reader user cannot complete a
purchase.
One thing we tested and disproved: we expected a site's platform to predict its defects. Across ten Shopware shops, the mean overlap between any two sites' failure sets was 0.21, against 0.19 for unrelated sites, which is indistinguishable. Platform tells you almost nothing. Every site has to be measured.
Read this before you buy
Automated testing cannot establish conformance, and we will never tell you otherwise.
Two different measurements get confused here, so to be exact about both. Deque's study of 13,000 pages found automation detected roughly 57% of the individual defects in that dataset, counted by volume. Separately, only 16 of the 50 Level AA success criteria can be meaningfully automated at all, so 34 criteria, about 68%, require a human regardless of how many defects a scanner finds.
The volume figure is high because the defects machines catch best, missing alt text, low contrast, unlabelled fields, are also the most numerous. It is not a statement about how much of the standard is covered, and a vendor quoting it as though it were is misleading you.
What no tool can decide, including ours:
- Whether alternative text is genuinely descriptive, rather than merely present
- Whether keyboard focus order follows a logical reading sequence
- Whether an error message explains how to recover
- Whether captions are accurate to the media
- Whether a page is operable end to end with a screen reader
So we sell evidence, remediation and a documented process, never a compliance badge. In January 2025 the US Federal Trade Commission fined an overlay vendor $1,000,000 for claiming its automated product could make sites WCAG compliant; the final order bars it from repeating the claim without evidence. We think that claim was always dishonest, and the 34 criteria we cannot reach are where your agency bills its own hours on top of ours.
Pricing
Billed once a year. Cancel whenever, unused months back.
Agency
$1,990 / year
- Up to 15 client sites
- Weekly scans, light and dark schemes
- Playwright suite and computed fixes
- Reports white-labelled to your brand
- Dated remediation log per site
Agency 40
$3,990 / year
- Up to 40 client sites
- Everything in Agency
- More pages per scan
- Priority on new checks
One-off audit
$249 once
- One site, scanned deeply
- Full suite and remediation set
- No subscription
- Credited against a plan within 60 days
Why 15 sites for $1,990
That is $11.06 per site per month. accessiBe and UserWay charge $49 to $199 per site, per month, so fifteen sites with them costs $8,820 to $35,820 a year. Siteimprove starts in the tens of thousands. We are cheaper because the work is automated and because nobody is paying for a sales team.
We are also not the cheapest way to get a scan, and we will not pretend otherwise. Pope Tech scans 25 pages free, with unlimited rescans and monthly emails. If a scan report is all you need, use it. What you do not get there is a runnable test suite, computed replacement values, or a report carrying your own branding, which is the entire reason this exists.
Prices are in US dollars and exclude VAT. There is no card checkout. You email us, we agree the scope, and we send an invoice payable by bank transfer in euros to a European account, which is an ordinary SEPA payment on your side. Sales to VAT-registered businesses in the EU are invoiced without VAT under the reverse charge, Article 196 of Directive 2006/112/EC, so we will ask for your VAT number and registered address before invoicing. How ordering works, step by step.
What we do not do, as of today
No deploy-triggered scans (scans run on a schedule), no generated accessibility statement, no customer-facing dashboard, and no visual-regression diffing. We would rather list these than let you discover them after paying. Ask if any of them is the thing you actually need.
Who you are dealing with
A small operation, and we would rather say so.
conformance.report is a largely automated pipeline. It does the scanning, the evidence capture, the remediation and the reporting. We are new, and we have no logo wall to show you.
What we offer instead is verifiable: send us a URL and we will send back real defects in your own code, with the exact fixes, before you pay anything. That is a stronger claim than a customer list, because you can check it in an afternoon.
This site is scanned by the engine it sells, in both light and dark colour schemes, before it ships. It currently reports zero violations, and we publish our own accessibility statement saying why that still is not conformance. An accessibility vendor whose own site fails is not worth your money.