Nine in ten
Austrian sites fail

Short version: of 253 Austrian websites we could measure, 230 had at least one failure against WCAG 2.1 level AA — the standard the Barrierefreiheitsgesetz points to. That is 90.9%. The sample was deliberately biased towards the companies most likely to pass.

Why we ran this

The Barrierefreiheitsgesetz took effect on 28 June 2025. It obliges companies above the micro-enterprise threshold to make covered digital services accessible: online shops, booking and appointment tools, online banking, telecoms, e-books, ticketing. The standard is WCAG 2.1 level AA. Penalties reach €80,000. What compliance work actually involves, we describe on our accessibility and BaFG page.

The Österreichischer Behindertenrat reported that 2 of 100 tested sites met the requirements. We wanted to know what that looks like a year later, in segments we could define precisely and re-test.

What we tested, and why those three

The shops are the best case, on purpose. These are businesses that pay for a quality seal and are re-audited every year against criteria covering seller identification, contract terms, payment, returns and data protection. They are the most disciplined operators in Austrian e-commerce. If they fail, the wider field of more than 12,000 Austrian web shops is not doing better.

The pharmacies are not a sample at all. That segment is the complete official register — every pharmacy in Austria permitted to sell medicines at a distance. Coverage is 100%.

The numbers

Worst segmentShare of sites with at least one WCAG 2.1 AA failure

The hotel and pharmacy samples are small — 31 and 16 sites. We give them as counts rather than percentages on purpose, because a percentage on 16 sites suggests a precision that is not there.

What actually breaks

Share of certified online shops affected, n = 206

Two things stand out.

Nearly a third of mail-order pharmacies disable pinch zoom. A single attribute in the page header, usually added years ago to stop a layout wobbling on phones. The people most likely to enlarge text on a pharmacy website are the people least likely to be able to read it otherwise.

Links without an accessible name are worst in hotels — 67.7%. These are the icon links: the arrow that advances a gallery, the magnifier, the little globe that switches language. Visually obvious, and completely silent to a screen reader, which announces them as “link”.

How we measured

Each site was loaded once in a headless browser at 1440×900, given a moment to finish rendering, and scanned with axe-core restricted to the WCAG 2.1 A and AA rule sets. We excluded axe's best-practice rules deliberately: they are good advice, but they are not what the law asks for, and including them would have inflated every number here.

We discarded 45 measurements: 25 sites did not respond in time, and 20 returned a bot wall, a geo-block, a maintenance notice or an error page instead of the site. Those were removed from the sample entirely rather than counted as clean — which matters, because counting them as passes would have improved the headline figure by several points for no reason.

What this number is not

It is also not a statement about legal liability. Companies below both 10 employees and €2 million in turnover are exempt as micro-enterprises, and we cannot tell company size from a homepage. Some of the sites we measured are not obliged to comply at all. This is a technical measurement against a technical standard.

And we are not naming anyone. The findings are published in aggregate.

What we would fix first

If you run one of these sites, the ordering is fairly consistent:

1. Remove the zoom restriction. One attribute. Minutes of work, and it is the single cruellest failure on the list.
2. Give every icon link and icon button a name. Usually one attribute per element, and it turns a silent interface into a usable one.
3. Fix contrast in your design tokens, not page by page. Contrast failures cluster because they come from a palette decision. Fix the palette and hundreds of elements resolve at once.
4. Then test with a keyboard. Everything above is what a machine can find. Whether someone can actually complete a purchase without a mouse is the question the law is really asking, and no scanner answers it — that is the part of our audit we do by hand.

Our own site

We ran the same audit against applepie.digital. It found a contrast failure of 1.97 in our own navigation, caused by two stacked opacity values — a mistake a scanner catches and a human eye does not. We fixed it, and we publish what is still open. It seemed a poor idea to write this article otherwise.