Browser user preferences and design-system adoption

EN 301 549 v4.1.1 introduced clause 9.7 — user preferences for web pages: a page must not block or explicitly override the modes in which the browser presents it according to the user’s own settings (colour filters, contrast, text size, pointer size, text cursor, reduced motion), unless that is essential to the content. The page below tracks two related questions across the government sites in this corpus:

  1. How many pages show evidence of respecting browser-set preferences — CSS media features like prefers-color-scheme, prefers-reduced-motion, prefers-contrast, forced-colors, and the color-scheme property, plus script-side theme switchers (matchMedia, data-theme).
  2. Which design systems those sites use — a curated registry of government design systems, brand identities, and their adoption signatures, because a design system that handles preference support correctly propagates it to every site that adopts it.

Stats as of 2026-11-09 21:18 UTC — last scan: 2026-11-05

Browser user preferences and design-system adoption

5,031 of 87,696 available pages scanned (5.7% coverage) 4,622 of 5,031 scanned pages were reachable (91.9%) 3,999 of 4,622 reachable pages had at least one stylesheet checked (CSS tier: 86.5%) 2,007 of 4,622 reachable pages show at least one user-preference signal (43.4%) Pages containing an explicit override (forced-color-adjust: none): 38 — a risk signal for manual review under EN 301 549 clause 9.7, not a confirmed failure Pages matching the HTML signatures of a curated design system: 11

Registry coverage: 4 of 22 curated design-system entries have confirmed adoption signatures (6 entries still need review). Systems without signatures cannot yet be measured for adoption.

Signals detected

Signal Pages % of reachable
Dark/light colour scheme (prefers-color-scheme) 211 4.6%
Declared colour scheme (color-scheme) 257 5.6%
Reduced motion (prefers-reduced-motion) 1,836 39.7%
Increased contrast (prefers-contrast) 123 2.7%
Reduced transparency (prefers-reduced-transparency) 2 0.0%
Forced colours / high contrast (forced-colors) 142 3.1%

Counts are pages where the signal appeared in the page HTML or a fetched stylesheet. Attribution (HTML vs first-party CSS vs third-party CSS) is in the JSON feature_sources field.

Design systems and adoption

Design system Country Type Signatures Pages matched
Austria Design System ⚠ Austria design_system not curated 0
Digitale Huisstijl Antwerpen Belgium brand_identity not curated 0
Stijlgids Stad Gent Belgium brand_identity not curated 0
Vlaanderen Design System Belgium design_system not curated 0
Design System České republiky Czechia design_system not curated 0
Designsystem.dk Denmark design_system not curated 0
Helsinki Design System ⚠ Finland design_system ✓ 0
Suomi.fi Design System Finland design_system not curated 0
Système de Design de l’État France design_system ✓ 11
Angie Design System Germany design_system not curated 0
Government Digital Services Design Guide ⚠ Greece design_system not curated 0
Government Design System (Hungary) Hungary design_system not curated 0
Accessibilité public.lu ⚠ Luxembourg accessibility_guide not curated 0
NL Design System ⚠ Netherlands design_system ✓ 0
DESY Design System Spain design_system not curated 0
Dintel Design System Spain design_system not curated 0
Sharing Sweden Brand Identity Sweden brand_identity not curated 0
GOV.UK Design System United Kingdom design_system ✓ 0
EU Web Guide Design System — design_system not curated 0
EUAA Design System — design_system not curated 0
European Commission Component Library — design_system not curated 0
zeroheight design systems documentation ⚠ — design_system not curated 0

Pages matched is the number of scanned pages whose HTML contains that system’s adoption signatures (class prefixes, asset paths). It measures footprint on the pages in this corpus only — systems are adopted outside these domains too, and an empty signature list means adoption is not yet measurable for that entry.

⚠ = entry flagged for review (confidence below high or an unconfirmed classification). Registry source: data/design-systems.yaml.

User-preference signals by country

Country Scanned Available Reachable CSS checked Signal pages Signal % Dark scheme Reduced motion Forced colours Override risk Design-system pages
Austria 822 822 748 720 545 72.9% 23 540 21 0 3
Belgium 1,329 1,329 1,192 1,064 413 34.6% 77 376 41 6 3
Bulgaria 353 353 295 223 111 37.6% 2 111 5 0 0
Canada 2,527 4,469 2,387 1,992 938 39.3% 109 809 75 32 5
Total 5,031 87,696 4,622 3,999 2,007 43.4% 211 1,836 142 38 11

How to read these numbers

Signals are collected from the page HTML (inline styles, <meta name="color-scheme">, matchMedia scripts) and from fetched stylesheets (capped per page; shared CSS is cached). They are evidence of effort, not EN 301 549 clause 9.7 conformance: a page with no media query can still comply, and a page with one can still break in forced-colours mode. Only manual testing can clear a site.

📥 Machine-readable results: user-preferences-data.json (by_url carries one row per scanned URL so every aggregate above can be independently recomputed) and user-preferences-data.csv (same rows, one per URL, boolean columns per signal), both published as workflow artifacts.

Context: EN 301 549 clause 9.7 — user preferences for web pages.

What this scan measures (and what it does not)

These are signals of effort, not conformance verdicts.

  • Clause 9.7 is a do not override rule. A page with no media queries can still comply, and a page whose stylesheet mentions prefers-color-scheme can still break in forced-colours mode. Only manual testing with the real settings (contrast themes, reduce motion, dark mode, 200% zoom) can clear a site.
  • Detection runs in two tiers: the page HTML (inline styles, <meta name="color-scheme">, inline scripts), and the page’s linked stylesheets (capped per page, cached across the run). Pages whose stylesheets could not be fetched are reported separately — a “signal %” computed over a corpus where the CSS tier never ran would understate support, so the CSS-tier coverage is published alongside the counts.
  • forced-color-adjust: none is recorded as an override risk because clause 9.7’s NOTE 4 names that property as the way a page overrides user settings. It can be legitimate (a colour swatch where the colour is the information); it flags a page for manual review, it does not fail it.
  • Design-system adoption is matched from curated signatures (data/design-systems.yaml). Entries with no confirmed signature cannot be measured for adoption — that gap is published as registry coverage rather than reported as zero adoption.

Machine-readable data

  • user-preferences-data.json — summary, by_country, design_system_registry, and a by_url array with one row per scanned URL (features, per-origin feature_sources, override signals, matched design systems). Every aggregate above is recomputable by grouping these rows.
  • user-preferences-data.csv — the same rows as a flat file with one boolean column per signal, UTF-8 BOM for spreadsheets.

Both files are uploaded as GitHub Actions workflow artifacts by the generate-scan-progress workflow, not committed to the repository.