On 19 August 2026 we ran our deterministic accessibility engine over the homepages of 191 Czech consumer-facing websites: online shops, banks and insurers, transport and travel operators, telecoms and the automotive sector. Of the 145 sites that could be assessed, the median score was 24 of 100, and 112 sites, 77 percent, graded F. Only 3 sites, 2 percent, had zero detectable violations. The single most common failure is insufficient colour contrast, present on 94 percent of assessed sites.
| Key result | Value |
|---|---|
| Sites targeted | 191 |
| Sites assessed | 145 (46 blocked the scanner or failed to render) |
| Median score (0 to 100) | 24.2 |
| Share graded F (score below 40) | 77.2 % |
| Share graded A (score 90 or above) | 2.1 % |
| Sites with zero automated violations | 3 (2.1 %) |
| Most common failure | Colour contrast: 136 sites, 3,522 instances |
These are automated results, which makes the failure count a floor, not a ceiling: automated testing detects only a subset of the WCAG requirements that Czech Act No. 424/2023 Sb. points to, so a site failing here fails before the harder, manual criteria are even considered.
What we scanned and how
The engine drives axe-core 4.12.1 through a real browser (Playwright) against each site's homepage, computes a 0 to 100 score weighted by violation impact and instance count, and maps it to grades (A from 90, B from 75, C from 60, D from 40, F below). The method is deterministic: no AI, no sampling, the same input produces the same output. The run manifest, including engine versions and the ruleset hash, is part of the published dataset. We used the same engine, the same ruleset and the same method as July's German benchmark, so the two sets of numbers compare directly.
Three honesty rules apply throughout. Sites that blocked the scanner (29) or failed to render (17) are reported as not assessed, never counted as clean. Bot walls are never bypassed. And we publish aggregates only: no individual site is named, because a homepage snapshot from one date should not follow a company around as a public verdict.
The five failures that dominate
None of the most common failures is exotic; all have been in WCAG 2.1 AA since 2018 or earlier.
| Failure (axe rule) | Sites affected | Share of assessed | Instances | Impact | WCAG criterion |
|---|---|---|---|---|---|
Colour contrast (color-contrast) | 136 | 93.8 % | 3,522 | serious | 1.4.3 |
Links without a name (link-name) | 71 | 49.0 % | 633 | serious | 2.4.4 |
Images without alt text (image-alt) | 36 | 24.8 % | 512 | critical | 1.1.1 |
Buttons without a name (button-name) | 30 | 20.7 % | 183 | critical | 4.1.2 |
Missing zoom-capable meta viewport (meta-viewport) | 21 | 14.5 % | 23 | moderate | 1.4.4 |
Colour contrast is a design-system problem: 3,522 instances across 136 sites means the failing grey-on-white text is baked into the theme, not into one page. Links and buttons without an accessible name are worse in practice than their counts suggest, because they leave screen reader users hearing "link" and "button" with no clue what either does. Images without alt text on a shop homepage typically means product images, so the products themselves are invisible to assistive technology.
The comparison with Germany: Czech banks are not the exception
In July we ran the same method over 390 German consumer websites. The comparison is unflattering: the German median was 38, the Czech median is 24. In Germany 51 percent of assessed sites failed; in Czechia it is 77 percent. Three quarters of Czech sites do not even reach the German median: the Czech 75th percentile (38.1) equals it exactly.
The sharpest difference is banking. German banks and insurers were the bright exception, median 76 with only 11 percent graded F. Czech banks and insurers sit at median 27 with 77 percent graded F, the same failure share as the online shops.
| Sector (CZ) | Sites assessed | Median score | Share graded F |
|---|---|---|---|
| Telecom * | 9 | 34.8 | 66.7 % |
| Banking and insurance | 26 | 27.4 | 76.9 % |
| Automotive * | 8 | 24.0 | 87.5 % |
| E-commerce | 82 | 20.8 | 76.8 % |
| Transport and travel | 20 | 20.4 | 80.0 % |
The starred rows are small samples: read telecom and automotive as indicative only.
For the German banks we attributed the gap to the financial sector's long regulatory exposure. The Czech numbers qualify that explanation: the mere existence of regulation is not enough, what matters is how long and how hard it has actually pressed on the sector. Nothing comparable is visibly pressing on Czech finance yet.
What this means under Act No. 424/2023 Sb.
Act No. 424/2023 Sb., the Czech transposition of the European Accessibility Act, has been effective since 28 June 2025. E-commerce services are explicitly in scope (§ 2(2)(d)), the service provider's duties are set by § 13(1), and Annex 1, Section III, point 3 requires websites to be perceivable, operable, understandable and robust, the same four principles the WCAG criteria above implement. Enforcement sits primarily with the Czech Trade Inspection Authority (ČOI), with the Czech Telecommunication Office (ČTÚ) covering electronic communications services. Fines run up to CZK 5 million (source: Act No. 424/2023 Sb.). Only micro-enterprises providing services are exempt: fewer than 10 staff and annual turnover or balance sheet total of 2 million euro or less.
Against that backdrop, the numbers above describe a market where, one year into the obligation, the median covered homepage fails checks that catch only the cheapest defects. Contrast, link names and alt text are precisely the findings a market surveillance authority, or a competitor's lawyer, can verify with free tooling in minutes.
If your site is in the failing three quarters, the fixes are unusually cheap for a compliance obligation: contrast is a stylesheet change, link and button names are markup, alt text is content work. Our six-step guide to becoming audit-ready sets out the order we recommend, and a free scan shows in minutes where a specific site stands on exactly the checks used in this benchmark.
What this data does not say
Automated results are an upper bound on conformance, not proof of it. axe-core evaluates the machine-checkable subset of WCAG 2.1 AA; keyboard traps, focus order, form error handling and cognitive load need human testing, so a site that passes every automated check can still fail a manual audit. We scanned homepages only, on one date, so a checkout flow may be better or worse than the front door. And a score of 100 here means "nothing automated found", not "accessible".
The reverse reading is the defensible one: a site that fails these checks demonstrably fails WCAG 2.1 AA, because every rule in the tables above maps to a specific criterion.
One more honesty note: the scanner now also detects each site's platform, but our target list was deliberately composed of large brands, which mostly run custom stacks. This data therefore says nothing about platforms like Shoptet that dominate among smaller Czech shops, and we publish no platform comparison.
The dataset
The full aggregate dataset behind this article, population counts, score distribution, grade shares, sector statistics, rule frequencies and the run manifest with engine versions and ruleset hash, is published as open data: accessibility benchmark dataset, Czechia, August 2026. Cite it freely with attribution to Inkluso. Per-site results are not published; the aggregates are the story.
We plan to repeat this benchmark for the Polish market, where the ustawa of 26 April 2024 raises the same questions, and to re-run both the Czech and the German scans over time to track whether enforcement moves the numbers.
See where your site stands
A free scan checks your site against WCAG 2.1 AA and shows the first findings in minutes.