Accessibility statement.

The standard we build to

We build and maintain this site — and every site we make — against the Web Content Accessibility Guidelines, version 2.1, at Level AA. Those guidelines are written by the World Wide Web Consortium, the body that sets the technical standards of the web, and they rest on four ideas: that content must be perceivable, operable, understandable and robust. At Level A and Level AA together, a page is measured against 50 separate requirements.

A newer version, 2.2, was published in October 2023 and adds nine further requirements. It does not replace 2.1 or make it obsolete; the two share the same foundation, and anything that meets 2.2 also meets 2.1. We name 2.1 Level AA because that is the version we actually measure against every month, and we would rather publish the standard we test than the newest one on the shelf.

What we do when we build a page

Accessibility is part of the build, not an overlay bolted on afterwards. Overlay widgets and accessibility badges do not make a site usable with a screen reader — a correct document does.

• Real headings in the right order and proper landmark regions on every page
• Keyboard operable throughout, with a visible focus indicator
• Contrast measured against the 4.5:1 and 3:1 thresholds, never judged by eye
• Text alternatives on meaningful images; decorative images hidden from assistive tech
• Links in body copy underlined, so they do not rely on color alone
• Every page usable down to a phone-sized screen without sideways scrolling

How this site is tested

Once a month every page on this site is scanned with axe-core, an open-source accessibility testing engine. The scan runs 62 rules drawn from the Level A and Level AA rule set, and it runs on each page twice: once at a desktop screen width and once at a phone width of 375 pixels, because a problem can exist at one size and not the other. Each page is sampled twice after it loads and then loaded again as a returning visitor. Anything the scan flags is checked a third time in a clean browser session before it is recorded.

What an automated scan cannot tell you

An automated scan is a measurement, not a verdict. Research published by Deque, the company behind axe-core, put automated coverage at about 57 percent of accessibility problems — measured across more than two thousand audits and roughly three hundred thousand individual issues. The rest need a person. A scan can confirm that an image carries alternative text; it cannot tell you whether that text describes the picture. It can confirm that a heading exists; it cannot tell you whether the wording is clear.

The engine also returns results it cannot decide either way, which it marks as needing review, and each report says how many it found. Those are neither failures nor passes — they are the places a human review would look first.

We have not published a keyboard-only or screen-reader review of this site, and a passing scan is not a substitute for one. We do not claim this site satisfies every requirement. What we publish is what was actually measured, on the dates it was measured.

If something here does not work for you

Tell us, and we will help you directly while we fix it. Email us and, if you can, say which page you were on and what happened — but you do not need to know the technical reason, and you will never be asked to explain your disability or prove anything. We can also provide anything published on this site in another format.