Tabpath Research · 22 September 2026 · Note 04

48 of 48

Every storefront we measured that was paying for an accessibility overlay still failed automated WCAG checks. Not most of them. All of them.

Sample 1,161 storefronts scanned, 48 running an overlay Method axe-core 4.13 · WCAG 2.0/2.1/2.2 A + AA Scanner corrected for the scroll-reveal artifact Re-verified live, days later

48of 48 overlay storefronts still failing

0came back clean

29median failing instances per storefront

4–145range of failing instances

An accessibility overlay is a script you pay to add to your site. It promises to make the site accessible without changing the site. We scanned 1,161 e-commerce storefronts. Forty-eight were running one. Every single one still failed automated accessibility checks, with a median of 29 failing instances and six or seven distinct rule failures each.

What we measured

Our census ran a full scan against 1,161 reachable storefronts. The scanner records which accessibility overlay, if any, a page loads, by matching the vendor's own script signature. It then runs axe-core 4.13 against WCAG 2.0, 2.1 and 2.2 at levels A and AA.

Two details matter for whether you should believe the number. First, the scanner waits out a generous settle window, because overlay vendors inject their changes after page load and judging them too early would be unfair to the overlay. Second, this census ran after we fixed the scroll-reveal artifact described in Phantom Contrast — a bug that had inflated our own contrast counts by as much as 24×. Both corrections push the numbers down, in the overlay's favour.

Overlay vendor	Storefronts	Still failing
accessiBe	31	31
UserWay	9	9
AudioEye	4	4
EqualWeb	4	4

Failures ranged from 4 instances on the cleanest site to 145 on the worst, across 2 to 15 distinct rules. The median storefront failed six or seven separate rules.

What actually fails

These are the rules that recurred most across the 48. The count is the number of storefronts affected.

axe rule	Sites	What it means
region	45	Content sits outside any landmark, so screen-reader users cannot skip to it
color-contrast	33	Text below the WCAG contrast minimum against its background
link-name	31	A link with no accessible name — a screen reader announces nothing useful
landmark-unique	27	Duplicate landmarks that cannot be told apart
image-alt	18	Images with no alternative text
button-name	15	A button with no accessible name
heading-order	15	Heading levels skip, breaking document navigation
nested-interactive	11	Controls inside controls, which assistive technology cannot reliably operate

Why we trust the link-name number in particular

In note 03 we published our own 94% false-positive rate on missing-name findings: our first-pass check produced 31 and only two survived. The check that got it right in that exercise was axe-core's own name rules, which flagged none of the 29 false ones. These counts come from axe, not from the naive check we retired. That is exactly why we retired it.

We re-ran it live to be sure

A census is a snapshot, and a snapshot can be wrong. So we took eight of the 48 and re-scanned them live, days later, from a clean browser. All eight reproduced — the same overlay still installed, the same failures, in two cases to the exact instance. Census figure first.

Storefront	Overlay	Census	Live re-scan
A — US food brand	UserWay	145	144
B — US beverage brand	UserWay	118	122
C — US apparel manufacturer	accessiBe	115	116
D — US pet products	accessiBe	101	97
E — US consumer brand	accessiBe	96	96
F — US furniture retailer	accessiBe	70	80
G — US art supplies	EqualWeb	58	59
H — US retailer	UserWay	49	49

The regulator has already ruled on the marketing

On 22 April 2025 the Federal Trade Commission approved a final order requiring accessiBe to pay $1 million. The FTC alleged the company's claims were false: that accessWidget "did not make all user websites WCAG-compliant and these claims were false, misleading, or unsubstantiated," and that the company "deceptively formatted third-party articles and reviews to appear as if they were independent opinions." That is the regulator's finding, not ours — it is published here.

Our own measurement is narrower and duller than the FTC's case. We are not saying an overlay does nothing. We are saying that on every storefront we measured, buying one did not produce a site that passes automated checks — the cheapest and most basic bar there is.

What we are not claiming

Automated testing reaches roughly a third of WCAG. These counts are a floor, not a verdict. A site with zero automated failures can still be unusable with a keyboard or a screen reader.
We are not saying the overlay caused these failures. The failures are in the underlying site. The claim on trial is whether paying for an overlay removes them. On this sample, it did not.
This is not a legal opinion. Failing an axe rule is a technical test result. Whether any site complies with any law is a matter for lawyers and courts, not for us.
We excluded target-size findings, even though axe flagged them on 16 of the 48, because we have not implemented the spacing exception that WCAG 2.5.8 allows. We said publicly that we would withhold that category until we do, so we are withholding it.
Overlay detection is signature-based. A vendor that renamed its script would be missed, so 48 is a floor for how many of the 1,161 are running one.
Storefront identities are withheld, as a matter of policy. This note is about vendors' claims, not about naming merchants who bought in good faith.

Find out what your overlay is actually doing

If you pay for one of these widgets, you can have the same measurement run on your own site. Send us the URL. We will run the scan and send you the result: which overlay we detect, what still fails, and which findings we could not determine. No charge, and no obligation to buy anything afterwards.

We will also tell you if the answer is boring. Some sites come back with very little, and we will say so.

Email us your URL [email protected] · usually back within two business days · we will not add you to a mailing list