OPEN RESEARCH
Europe: municipal websites in Germany and Spain
What a keyboard-only visitor meets on the website of their own town hall. Traversed with the same method as the U.S. study: a real Tab keypress through the browser driver, eight pages per site, a screenshot kept as evidence.
| Outcome | Records | Share of measured |
|---|---|---|
| Formally open, but a barrier in practice | 11,801 | 77.8 % |
| Partial focus or labelling barrier | 1,846 | 12.2 % |
| Reachable by keyboard | 1,516 | 10.0 % |
| Opened, but could not be checked as a human would | 3,022 | — |
| Domain does not exist | 858 | — |
| Server does not answer | 550 | — |
| Closed to checking (403) | 446 | — |
| Broken encryption | 443 | — |
| Server-side error | 351 | — |
What the shares are taken from
Shares are taken from MEASURED records, not from all of them. “Could not be measured” covers six different outcomes where there was nothing to measure: no domain, silent server, page closed to robots, broken encryption, server error, and a page that opened but resisted checking. Together they are 27.2 % of everything collected. Computed against all records, every figure would drop by roughly a quarter — and would describe the share of dead addresses rather than accessibility.
Compared with the U.S. — same denominator
Among measured records, 10.0 % of European municipal sites were reachable by keyboard against 25.4 % of American ones. Both numbers are computed the same way. Any other comparison would be a substitution: in the U.S. the “could not be measured” share is twice as high (50.5 % against 27.2 %), so counting against all records would make Europe look better — not because it is more accessible, but because fewer of its domains are dead.
What we withdrew, and why
We hold a second European dataset — 71,065 records across ten countries. It is not published. The engine behind it tested the keyboard with a synthetic dispatchEvent: such an event does not move focus, because the browser performs the default action only for a real keypress. Focus therefore never moved, and the verdict was pronounced on a state that stood still. Verified on our own site: this method would have declared aifa.works inaccessible. Ten countries would look more impressive than two — but a figure produced by a broken instrument does not become true by being large.
71,065 · 10 · dispatchEvent
What this study does not say
- — Two countries, not Europe. Germany and Spain are what a real keypress measured; the rest are being re-measured.
- — The keyboard is not all of accessibility. Someone using a screen reader or magnification will meet other obstacles, and those are not measured here.
- — Eight pages per site, not the whole site. Home, payment, contact and five more common types — the places where people get stuck most often.
- — A machine does not replace a person. It finds an obstacle; it does not say whether there is another way round it.
Who did this
Maksim Galatin, self-funded research. The scanner was written from scratch; raw data and methodology are published in full.
Conflict of interest disclosed: the author sells accessibility remediation services. The research is self-funded; nobody paid to be included in the sample or excluded from it.