Data: HTTP Archive, crawl of 1 September 2026, mobile, home pages only, automated checks only. Aggregates only; no store is named. Not legal advice.

Read these three limits first

Automated checks only. Automated tools detect only part of the barriers. A page with no failing audit is not an accessible page.
Home pages only. Product pages, cart and checkout are not in this data.
A sample, one day. 5,463 Shopify home pages on EU country domains and .eu within the Chrome UX Report top 1 million, one crawl on 1 September 2026, mobile emulation.

Summary

Since 28 June 2025, online shops that sell to consumers in the EU must meet the accessibility requirements of the European Accessibility Act (Directive (EU) 2019/882), as transposed into 27 national laws. Microenterprises that provide services are exempt. We looked at what public data says about Shopify stores in the EU fifteen months later.

We analysed the accessibility audits that the public HTTP Archive ran with Lighthouse (which uses the axe-core rules) on 5,569 Shopify home pages on EU country domains and .eu. 5,463 pages had a Lighthouse result. Five findings:

94.2% of home pages failed at least one automated check that maps to a WCAG 2.1 level A or AA success criterion (5,148 of 5,463). Only 3.5% (190 pages) had no failing audit among the 36 checks we counted. Having no failing automated check does not mean a page is accessible: most WCAG criteria need a human to test them.
Colour contrast is the most common barrier: 74.6% of home pages had text with insufficient contrast (WCAG 1.4.3; 95% CI 73.4–75.7%).
Assistive technology often cannot name what it finds: 54.3% of pages had links without a discernible name, 36.0% had images without an alt attribute, 18.8% had buttons without a name.
17.0% of home pages block zooming through the viewport settings (WCAG 1.4.4). On .de domains the share was 23.9%, the highest of any country group.
Overlays are rare and do not change the picture. An accessibility overlay or toolbar product was detected on 54 of 5,569 home pages (1.0%). On pages with an overlay, the automated checks still found at least one WCAG 2.1 A/AA barrier on 88.5%, and the average number of failing checks was the same as without (3.3).

Two further patterns: differences between countries are small (median Lighthouse accessibility score between 0.87 and 0.90 in every country group with at least 50 pages), and pages that load more third-party technologies fail more checks (2.7 failing checks on average with fewer than 20 detected technologies, 4.1 with 40 or more).

These numbers describe only what a machine can detect on a home page. They are a lower bound, not a measure of how usable these stores are for disabled people.

Why this matters

The EAA requires e-commerce services to be “perceivable, operable, understandable and robust” (Annex I), and requires providers to publish information on how their service meets the requirements (Art. 13 and Annex V). In Germany, the Barrierefreiheitsstärkungsgesetz (BFSG) adds that this information must name the market surveillance authority (BFSG Anlage 3 Nr. 1 d).

The technical reference for web content is the harmonised European standard EN 301 549. Version 3.2.1 incorporates WCAG 2.1 level AA. Version 4.1.1, which incorporates WCAG 2.2, was published by ETSI on 2 September 2026; according to AccessibleEU it is not yet the legal reference standard.

Regulators have started to look. In March 2026 the Dutch ACM reported that at 61% of the roughly 100 largest Dutch webshops it tested, it was impossible to place an order with assistive technology, and that a further 33% had serious problems.

What has been missing is a picture of the stores that make up most of the EU Shopify market: not the largest retailers, but the many mid-sized brands that run on a Shopify theme and a set of apps. Public crawl data lets us measure the part of the problem that machines can see.

Data and method

Source

The HTTP Archive crawls millions of websites every month and publishes the results as a public BigQuery dataset. Its site list comes from the Chrome UX Report (CrUX), grouped into popularity buckets. For each site it runs Lighthouse, Google’s open-source page audit tool. Lighthouse’s accessibility audits use the axe-core rules; each audit is pass or fail for the whole page.

We queried the table httparchive.crawl.pages for the crawl of 1 September 2026, mobile client, root pages only, where:

the technology detection (a fork of Wappalyzer) found Shopify,
the host ends in an EU member state country code or .eu,
the CrUX popularity rank is within the top 1 million (a cost limit: the full table would have exceeded the free BigQuery quota).

The query extracted, per page: the Lighthouse accessibility score, the list of failing audits among 36 axe-based audits (failing = audit score 0), the CrUX rank bucket and the detected technologies. Only aggregates were kept.

Sample

5,569 home pages (5,550 distinct registrable domains; some brands run regional subdomains).
5,463 had a Lighthouse result. 106 had none (Lighthouse did not complete) and are excluded from all audit statistics.
Largest country groups by domain: .de 1,231, .it 623, .fr 615, .nl 523, .ro 363, .es 302, .pl 272, .eu 254.
Popularity: 144 pages in the top 100k, 1,944 in 100k–500k, 3,481 in 500k–1M.

What we counted

Failing audit: a Lighthouse accessibility audit with score 0 on that page. One failing element is enough. We did not count how many elements failed.
WCAG 2.1 A/AA barrier: any failing audit whose axe-core rule maps to a WCAG 2.1 level A or AA success criterion. This excludes heading-order and tabindex (axe best practices) and target-size (WCAG 2.2 criterion 2.5.8, not in WCAG 2.1).
Overlay: a page where the HTTP Archive technology detection named one of these accessibility overlay or toolbar products: accessiBe, UserWay, EqualWeb, AudioEye, Nagich, Allyable, Accessibility Toolbar Plugin.
Country: the country code top-level domain of the page. This is the domain, not the merchant’s legal seat or customers.

Confidence intervals are 95% Wilson intervals. They describe sampling noise only, not the selection limits below.

All limits

Automated checks only. In a 2017 test by the UK Government Digital Service, ten tools together found 71% of 143 deliberately created barriers; the best single tool found 41%. The W3C states that evaluation tools “can not determine accessibility, they can only assist in doing so.”
Home page only. Product pages, cart and checkout, where the EAA explicitly requires accessible payment functions, are not in this data.
Mobile only. Lighthouse ran with mobile emulation.
One snapshot. One crawl, one day. Stores change their themes and apps often.
Not all EU Shopify stores. Stores on .com and other generic domains are excluded, as are stores outside the CrUX top 1 million.
Selected audits. We counted 36 axe-based audits. Lighthouse runs a few more.
Detection of technologies is imperfect. The overlay share is a lower bound.
Correlation, not causation. Where we compare groups, we cannot say what caused the difference.

Results

1. Almost every home page has at least one detectable barrier

Pages with at least one failing audit (5,463 pages with a Lighthouse result)
Measure	Pages	Share
At least one failing audit mapped to WCAG 2.1 A/AA	5,148	94.2%
Including WCAG 2.2 target size	5,220	95.6%
No failing audit among the 36 counted	190	3.5%

The typical page failed 3 audits (median; mean 3.3). The Lighthouse accessibility score had a median of 0.89 (quartiles 0.84 / 0.89 / 0.93); 176 pages (3.2%) scored 1.00. For comparison, the WebAIM Million 2026 report found detectable WCAG 2 failures on 95.9% of one million home pages worldwide, with a different tool and a different way of counting; only the direction is comparable.

2. The seven most common barriers

The seven most common failing audits
Lighthouse audit	What it means for a shopper	WCAG 2.1	EN 301 549 V3.2.1	Pages failing	Share (95% CI)
`color-contrast`	Text is hard to read for people with low vision	1.4.3 (AA)	9.1.4.3	4,076	74.6% (73.4–75.7)
`link-name`	A screen reader announces “link” with no destination, often icon links (cart, account, social)	2.4.4 (A), 4.1.2 (A)	9.2.4.4, 9.4.1.2	2,969	54.3% (53.0–55.7)
`target-size`	Tap targets too small or too close together	2.5.8 (WCAG 2.2 AA)	not in V3.2.1	2,032	37.2% (35.9–38.5)
`image-alt`	Images without an alt attribute, so their content or role is unknown	1.1.1 (A)	9.1.1.1	1,965	36.0% (34.7–37.3)
`heading-order`	Heading levels are skipped, which makes page navigation harder	best practice	none	1,881	34.4% (33.2–35.7)
`button-name`	Buttons (menu, close, carousel, quantity) without an accessible name	4.1.2 (A)	9.4.1.2	1,026	18.8% (17.8–19.8)
`meta-viewport`	The page prevents pinch-zoom	1.4.4 (AA)	9.1.4.4	930	17.0% (16.1–18.0)

All other audits failed on fewer than 9% of pages (full list: audits.csv). Most of these barriers sit in theme code or in theme settings: colour choices, icon links and buttons without a text label, the viewport tag, heading levels in theme sections. Missing alt text is often content (product and banner images) that merchants can fill in Shopify’s admin.

3. Small differences between countries

Country groups with at least 100 pages with a Lighthouse result
Domain	Pages	Median score	≥1 WCAG 2.1 barrier	Contrast	Link name	Image alt	Button name	Zoom blocked
All	5,463	0.89	94.2%	74.6%	54.3%	36.0%	18.8%	17.0%
.de	1,207	0.89	93.6%	75.1%	49.8%	34.2%	16.3%	23.9%
.it	609	0.89	92.3%	69.8%	53.2%	40.2%	16.3%	15.9%
.fr	608	0.88	94.4%	71.9%	56.6%	39.3%	19.9%	17.1%
.nl	513	0.88	95.3%	82.3%	53.6%	35.9%	21.4%	15.4%
.ro	354	0.88	93.8%	78.2%	61.6%	33.1%	18.4%	14.1%
.es	296	0.89	93.9%	72.3%	55.7%	40.5%	17.6%	14.2%
.pl	268	0.90	94.8%	74.6%	44.4%	35.1%	14.2%	11.2%
.eu	248	0.89	92.3%	68.1%	57.3%	37.9%	21.8%	15.7%
.dk	206	0.87	98.5%	74.8%	65.5%	37.4%	32.5%	18.4%
.ie	152	0.90	92.8%	69.1%	67.8%	28.9%	12.5%	20.4%
.fi	146	0.89	93.2%	76.0%	55.5%	29.5%	19.9%	18.5%
.se	136	0.885	93.4%	73.5%	55.9%	32.4%	25.7%	14.0%
.pt	110	0.88	95.5%	80.9%	59.1%	40.9%	21.8%	10.0%

Smaller groups are in by_cctld.csv. With 50–100 pages, differences of a few points are within noise. Two patterns stand out beyond noise: .de pages block zoom more often (23.9% vs 17.0% overall) and .nl pages fail colour contrast more often (82.3% vs 74.6%). We cannot tell from this data why.

4. Popularity makes little difference

Results by CrUX popularity bucket
CrUX rank bucket	Pages	Median score	≥1 WCAG 2.1 barrier	Mean failing audits
Top 100k	139	0.89	91.4%	3.45
100k–500k	1,907	0.89	94.5%	3.38
500k–1M	3,417	0.89	94.2%	3.27

5. More third-party technology, more failing checks

Results by number of technologies detected on the page
Detected technologies	Pages	Mean failing audits	Link name	Image alt	Button name	Zoom blocked
Under 20	1,139	2.69	43.5%	29.5%	12.5%	10.6%
20–29	2,567	3.26	52.8%	33.3%	17.6%	16.7%
30–39	1,492	3.74	63.0%	42.7%	24.4%	20.8%
40 or more	265	4.12	67.2%	51.3%	26.0%	26.0%

This is a correlation. Pages with more technologies are also often larger and more complex. But it matches what accessibility testers see in practice: widgets, pop-ups and embedded components bring their own unlabeled buttons and links. We do not attribute barriers to any single app from this data.

6. Overlays: rare, and no visible effect on detectable barriers

Home pages with and without a detected overlay
Measure	Overlay detected (52)	No overlay detected (5,411)
≥1 WCAG 2.1 A/AA barrier	88.5%	94.3%
Mean failing audits	3.29	3.31
Median Lighthouse score	0.87	0.89
Image alt missing	48.1%	35.9%
Button name missing	26.9%	18.7%
Colour contrast	59.6%	74.8%
Zoom blocked	11.5%	17.1%

Lighthouse audits the page after it has loaded, so any changes an overlay makes during page load are part of what was measured. With 52 pages, differences of this size are within noise in both directions. What the data does show is that nearly nine in ten of these pages still had barriers that an automated check finds. This is consistent with the joint statement of the European Disability Forum and the IAAP (2023), which says that overlays do not make a website accessible and do not meet European accessibility legislation. In 2025 the US Federal Trade Commission ordered one overlay vendor to pay $1 million, in part over unsupported claims that its product made websites meet WCAG.

What merchants and agencies can take from this

Start with the theme. Contrast, icon link and button names, the viewport tag and heading levels are theme-level fixes. One fix in a theme section repairs every page that uses it.
Fill in alt text. Product and banner images are content that merchants control in Shopify’s admin.
Review apps, not only the theme. More third-party components went with more failing checks.
Test beyond the home page and beyond automation. Checkout, product pages and keyboard and screen-reader use need manual testing.
Write down what you checked. The EAA asks providers to explain how their service meets the requirements (Annex V). A dated record of scans, fixes and manual tests is the basis for that explanation.

See which barriers an automated check finds on your own store, beyond the home page.

Request the free store check

Download the data

Aggregate CSV files only: no domain, URL or store name appears in them. The numbers can be reproduced from the public HTTP Archive dataset with the method above. When you use them, please cite “Testamur, EU Shopify accessibility 2026, HTTP Archive crawl of 1 September 2026”.

sample.csv: Sample: pages, ranks, country domains
audits.csv: Every counted audit: pages failing, share, 95% CI, WCAG mapping
by_cctld.csv: Results by country domain
by_rank.csv: Results by CrUX popularity bucket
by_technology_count.csv: Results by number of detected technologies
common_apps.csv: Most common third-party technologies
overlays.csv: Overlay products detected
overlay_vs_none.csv: Pages with and without an overlay
score_distribution.csv: Lighthouse accessibility score distribution
failing_audits_per_page.csv: Failing audits per page

About this study

This study was prepared by Testamur, a Shopify app in development, operated by Sieghart Michielsen (sole proprietorship, Antwerp, Belgium). Testamur did not scan any store for this study. All store-level data comes from the public HTTP Archive dataset, and only aggregates are published. No store is named or ranked. Corrections and questions: hello@testamur.app.

EU Shopify accessibility 2026: what automated checks find on 5,463 store home pages