Loading today’s stories
Build1 distinct publisher3 min readPublished
A scan of 149 top EU-domain sites found 64% failing Name, Role, Value. The readiness checklists on sale are for a standard that arrives about 17 months after the deadline these sites already missed.
The Engineer · Build desk
Compiled by The EngineerSomething wrong?How this is made
Six clause failures in a table do not mean six problems. A link wrapped around an icon with no accessible name fails Name, Role, Value and Link Purpose at the same time [7], so one missing attribute takes up two rows of the summary and one line in the repair queue. Conformance is claimed against clauses rather than rule counts [7], which is why the double entry is not a reporting error. It does mean the fix list is shorter than the table's arithmetic suggests.
The scale first: 64% of 149 sites [3][4] is about 95 sites [1] whose most frequent defect is an icon link, usually a magnifier or a basket, marked up with an image inside and no text anywhere [11]. The next most frequent is an advert [11]. Neither is architectural [11].
The second pass is where the sample gets thin, and the author says so. Underlying rules were recorded for 59 of the top 120 domains [8], just under half that list [2], and Name, Role, Value fails on 73% of that group [9] against 64% across the full sample [4], a nine point gap [3] the author attributes to the subsample sitting at the more-visited end [9]. Those shares are proportions within a worse group, not headline rates [9]. Inside it, every contrast failure came from a single rule and every non-text-content failure was an image with no alternative text [12].
The 34% target-size figure [15] arrives with its own caveat attached. Target size is a WCAG 2.2 criterion and is not currently required [16]; it appears at all only because it is the one of the six incoming criteria the rule engine can check [16]. A third of the sites showed a consent dialog to an anonymous visitor, and on one in five that dialog covered more than 40% of the screen [17]. The author declined to click accept at scale [17], so the number partly describes consent dialogs: 40% failing with one, 31% without [18]. Nine points [4], and the pages behind the dialogs are still not clean.
The finding that should reorganise a budget conversation is the median split. The more popular half of the measured sites is roughly three times less likely to come back clean [13], and the author, who cannot prove the mechanism from this data, points at third-party widgets and legacy code being shipped into one page by several teams [14]. Money is not the binding constraint [14]. Meanwhile 32 of the 149 sites failed nothing automated testing could detect, 21% [6], and the author is explicit that this is not the same as passing [6].
Which puts the checklist industry that has grown up around V4.1.1 [2] in an awkward spot. Five of the six criteria arriving with it, focus not obscured, dragging movements, consistent help, redundant entry and accessible authentication, were not measured because each is a judgement about behaviour that the page source does not contain [19]. The one that a scanner can see is the one that is not required yet [16]. The version cited on 30 November 2026 [1] lands roughly 17 months [5] after the deadline these sites already went past in June 2025 [5], failing a Level A criterion that has been in every version of WCAG since 2008 [4].
Ranked by verification strength, evidence, and original report placement.
The author scanned the most-visited websites on EU country domains, 149 sites in the full sample, and counted which clauses of EN 301 549 they fail under the version cited right now.
Sixty-four per cent of the sampled sites fail clause 9.4.1.2, Name, Role, Value. It is Level A, the lowest bar in the standard, and it has been in every version of WCAG since 2008.
EN 301 549 V4.1.1 is scheduled for citation on 30 November 2026. It swaps WCAG 2.1 for WCAG 2.2, and six new success criteria arrive at levels A and AA.
The version of EN 301 549 currently cited has been in force since before the European Accessibility Act deadline passed in June 2025.
Thirty-two sites of the 149, about one in five, failed nothing that automated testing can detect. The author notes that is not the same as passing.
The rule that most often breaks Name, Role, Value is a link with no accessible name, and the same defect also fails Link Purpose, so one missing label lands in two rows of the clause table. Conformance is claimed against clauses rather than rule counts.
Follow any of these and your For You feed starts watching them — no settings page required.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
One transparent but unreplicated self-reported scan
The methodology is stated plainly (149 most-visited EU country-domain sites, clause mapping against the currently cited EN 301 549, each page loaded twice with only double failures counted) and the author volunteers the limits that matter: subsample skew, clause double-counting, consent-dialog contamination and run-to-run instability up to 19%. That candour raises evidence quality above a bare claim. It is still one self-published source with no dataset, no site list, no ranking provenance and no independent replication, and the author's own reruns show individual clause results are not stable, so specific percentages should be read as approximate.
Standard already mandatory, field conformance thin
Adoption here is conformance with a standard that is already cited and backed by a passed regulatory deadline, not voluntary uptake of a technology. On the supplied measurement, uptake in the field is weak: 64% of 149 top EU-domain sites fail a Level A clause that has existed since 2008, 73% in the more-visited subsample, and only about 21% return no automatically detectable failure at all - a figure the author explicitly says is not the same as passing. The tooling side of adoption is also narrow: only one of the six criteria arriving in V4.1.1 has an automated check in the engine used.
Headline slightly outruns a noisy single scan
The article's core direction is well supported by its own data and it is more self-critical than most single-author studies, including a caveat before the target-size number and a disclosure of measurement flakiness. The overstatement is modest and mostly in framing: 'two thirds of Europe's biggest sites fail' is presented as a settled rate when it comes from one unreplicated scan whose reruns disagreed on 10-19% of sites, and the author says outright he would rather quote the flattering 6% instability figure. The piece simultaneously deflates a larger external hype - readiness checklists and 'WCAG 2.2 ready' scanners covering one criterion in six - which pushes the net gap only slightly positive.
No disclosed commercial stake, attention incentive present
The supplied source shows no vendor sponsorship, product pitch, tool sale or client relationship; the author criticises the readiness-checklist and scanner category rather than promoting an alternative, and volunteers findings that weaken his own numbers, which is against a pure attention incentive. Residual distortion risk is the ordinary one for self-published practitioner research: a striking headline rate drives reach, the underlying data is not released for checking, and the author states a preference for quoting the more flattering instability figure. Direction of any bias is therefore toward a cleaner, more quotable result than the data supports, not toward a commercial outcome.
Directionally credible, numerically soft
Confidence is moderate-low. The qualitative conclusions are the sturdiest part: the dominant failures are trivially fixable missing attributes, failures concentrate in single rules, the most-visited sites do worse, and automated tooling cannot assess five of the six criteria arriving in V4.1.1. Those survive the disclosed noise. Confidence in the exact percentages is materially lower given one publisher, no released dataset, a 59-of-120 measurement rate in the rule pass, and clause sets that changed on repeat runs hours apart. Regulatory facts about the citation date and the passed deadline are checkable in principle but are only asserted here.
build
The EAA exemption most vendors skip: Article 4(5) also cancels the paperwork1 distinct publisher
build
"Follow the same standard" is how mobile accessibility debt gets onto the books1 distinct publisher
build
A five-check gate is the only docs accessibility test that will survive contact with CI1 distinct publisher
build
Three manual interventions in a month, and every guard was working as designed1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
dev.to
1 article · August 27, 2026