Accessibility by shop system — what yours can do and what it cannot
No shop system is accessible out of the box, and none is hopeless. What matters is what the core provides, what the theme makes of it, and what your own content tears down again afterwards.
Three layers where barriers appear
The core is what the vendor ships: forms, cart, checkout, default theme. This is where most has moved since 2024, because the vendors are under the same deadline pressure as their customers.
The theme is what your agency made of it. In our measurement this is where most findings sit — custom colour palettes, hand-built dropdown menus, icon buttons without a label.
The content is what gets added daily: product images without alternative text, promotional banners in light grey on white, a “more” link five times on the same page. This layer cannot be solved once; it needs a routine.
Update only the first layer and skip the other two, and you have a new system with the same barriers.
Shopware 6
In its developer documentation Shopware explicitly commits to WCAG 2.1 AA and BITV 2.0 and has been rebuilding the storefront step by step. The point that matters for operators: in 6.6 the improvements that change existing behaviour sit behind the feature flag ACCESSIBILITY_TWEAKS and are off by default. From 6.7 they are the default. Anyone running 6.6 without that flag simply does not have them.
In the same documentation Shopware lists open issues still being worked on. So a version jump is a good start and not a proof.
Our measurement included four Shopware shops, ranging from 7 to 152 findings — the widest spread of any system tested. Notably frequent: aria-required-children in three of four shops, i.e. hand-built menu and list structures whose ARIA roles do not fit together. That is typical of customised navigations and does not come from the core.
Shopify
With its newer default themes Shopify provides a decent foundation, and the hosted checkout is barely modifiable by operators anyway — which takes one of the trickiest areas out of your hands. The variance comes from store themes and from apps.
In our measurement the two Shopify shops came in at 5 and 55 findings. Leading the list: link-name and button-name — icon controls for cart, search and account that are visible but nameless to screen readers. Cheap to fix and almost always missed.
Be careful with apps promising accessibility “at the push of a button”. Those are usually overlays — more on that below.
WooCommerce and WordPress
WordPress brings usable semantics with the block editor and its default themes. The problem is combinatorics: theme plus page builder plus a dozen extensions, each injecting its own HTML. A contact form plugin that renders its labels as placeholder text opens a gap that is invisible in the theme.
The two shops measured came in at 27 and 35 findings, with label errors in forms and the usual contrast problem. So the practical lever on WordPress is not the theme but the plugin list: every extension that produces visible front end deserves its own check.
Two further WordPress shops could not be measured because their bot protection refused the test request. We do not circumvent such blocks — that goes into the report, and it is more honest than inventing a number.
JTL-Shop, OXID, Magento, TYPO3 and the site builders
JTL-Shop 5 with the NOVA theme is in the same position as Shopware: the vendor is improving, but every theme customisation is your responsibility. There was no JTL shop in this sample, so we quote no number for it.
OXID eShop stood out across three shops measured (13 to 54 findings) with nested interactive controls — clickable elements inside other clickable elements, which become unreachable by keyboard. Plus SVG icons without a text alternative.
Magento / Adobe Commerce performed best across two shops (2 and 13 findings). That says little about the system and a lot about those two operators — with two measurements, anything else is tea leaves.
TYPO3 was represented by a single site, and at 157 findings it topped the entire run. That too is not a verdict on TYPO3, but on this one site.
Site builders such as Wix, Jimdo or IONOS have a problem of their own: you can only change what the builder allows. Contrast and alternative text usually work, but if a component emits faulty HTML your only options are to drop the component or change provider. Check that before you rebuild.
What we actually measured
So it is clear what the numbers above rest on — and where they stop.
On 25 August 2026 we approached 16 German-language sites whose system could be identified unambiguously from their source. 14 of them could be tested; two refused access. Up to six pages each were tested — home, category, product, cart, contact, legal notice — against the automatically testable criteria of EN 301 549 (WCAG 2.1 AA).
| System | Shops | Pages | Findings per shop | Most common finding |
|---|---|---|---|---|
| Shopware 6 | 4 | 19 | 7 – 152 | contrast, ARIA structure |
| Shopify | 2 | 12 | 5 – 55 | contrast, nameless links |
| OXID eShop | 3 | 13 | 13 – 54 | contrast, nested controls |
| WooCommerce | 2 | 10 | 27 – 35 | contrast, missing form labels |
| Magento | 2 | 11 | 2 – 13 | contrast, images without alt text |
| TYPO3 | 1 | 5 | 157 | contrast, ARIA structure |
Total: 70 pages, 654 consolidated findings out of 1,553 raw hits, 2,025 passed checks. A sample of this size shows spread and recurring patterns — it is not a representative study and not a ranking of systems.
The clearest result: contrast errors in 14 out of 14 shops tested. Across every system, regardless of size and budget. The second most common finding was link-name in 7 of 14 — links without a recognisable name, almost always icon controls.
And the more important limitation: an automated test covers only part of the criteria. Keyboard operability of the ordering process, focus visibility, reading order and the factual correctness of alternative texts appear in none of these numbers.
Why an overlay plugin does not satisfy the obligation
For almost every system there are extensions that add a toolbar: larger text, higher contrast, a read-aloud function. They install quickly and change nothing about the underlying HTML.
The German surveillance bodies consider such tools unsuitable for meeting the requirements, and the German Federation of the Blind and Partially Sighted warns against these vendors' marketing claims. People who use screen readers already have one — a second, bolted-on read-aloud function tends to interfere rather than help.
An overlay is therefore not merely ineffective; it is a visible signal that someone took the shortcut. That is not a good opening for an inspection.
What remediation costs
Four cost blocks, current market prices and why the ranges are so wide.
Does this apply to me at all?
Microenterprises are exempt for services. Three questions are enough to place yourself.
Accessibility statement
A free generator for the text the law requires separately from the technology.
Enforcement and market surveillance
How the procedure really runs — and why the loudest route is not the likeliest.
What is your number?
The figures above are other people's shops. Your own findings are one address and less than a minute away.