Version 1.0 Last revised: August 17, 2026
The Known Good wants its data and analysis to be usable by as many people as possible, including people who use assistive technology. This statement explains how far the site currently meets recognised accessibility guidelines, what we know does not yet work, and how to reach us.
The Known Good is an independent, non-commercial personal project. It has no user accounts, no logins, and nothing is for sale — the site is free to read and the data is free to download.
Scope
This statement covers the website at https://theknowngood.com and its subpages.
It does not cover our blog at https://blog.theknowngood.com. The blog is delivered through a
third-party publishing platform (Hashnode) whose page templates and rendering are not under our
control, and it is outside the scope of this evaluation.
Standard we aim for
We aim to meet the Web Content Accessibility Guidelines (WCAG) 2.2, level AA.
Conformance status
The site is partially conformant with WCAG 2.2 level AA. "Partially conformant" means that some parts of the content do not fully conform to the standard. The specific shortfalls we are aware of are listed under Known limitations below.
Feedback — tell us about a barrier
If you find something on the site you cannot use, please tell us and we will do our best to put it right.
- Email: [email protected]
- We aim to respond within 10 business days.
Reaching us this way means sending us your email address and whatever you include in your message; we handle that as ordinary correspondence, as described in our Privacy Policy.
An accessibility report you send us is treated as correspondence under our Privacy Policy, not as a "Submission" under section 9 of our Terms of Use, so we claim no licence over the content of your report.
How we evaluated the site
This was a self-evaluation. We tested with the axe-core 4.13.0 automated rule set driven through Puppeteer 25.8.0 using Chrome for Testing 152, on Node.js 26, combined with manual keyboard, focus, reflow, and interaction checks. We audited a representative subset of page types — the home page, leaderboards, a model detail page, a model comparison, an evaluation page, a provider page, the recommender, the data-downloads page, and the policy and information pages — each in light and dark themes at desktop and mobile widths, and generalised the findings by page type. Not every individual page was tested.
Testing was carried out against WCAG 2.2. Under WCAG 2.2, Success Criterion 4.1.1 (Parsing) no longer applies and was not assessed.
The evaluation spanned three builds of the site: the initial audit (build 20260817T212023Z); a
re-check after a server configuration fix that corrected how the data-downloads page is served
(build 20260817T222030Z); and the final measurements after a stylesheet change that improved
text contrast (build 20260818T011408Z). A server configuration fix and a stylesheet change were
therefore made between the initial audit and the final measurements. These build identifiers are UTC
timestamps — which is why the final one falls on the following calendar day — whereas the dates given
elsewhere in this statement are local.
The messages the newsletter form shows after submission were checked against a locally simulated server response, not a live sign-up.
Known limitations
We are aware of the following barriers. Where a workaround exists today, it is noted.
- Text and control contrast (partially remediated). We have improved the contrast of secondary and muted text throughout the site, in both light and dark themes. Some elements still fall short: the blue used for links and primary buttons, and the number labels printed on some coloured chart bars. For those chart labels, the same figures are available as text — a keyboard-accessible CSV download is offered for each chart, and evaluation and provider pages also show the full data as a table.
- Small targets. Some links and controls — particularly the labels beneath charts — are small and closely spaced, which can be difficult for people with limited dexterity or using a touch screen.
- Links shown by colour. In some blocks of text, links are distinguished from the surrounding text by colour alone, without an additional visual cue.
- Keyboard operation of two menus. The chart "Top N" count menu and the model-comparison picker can be opened with the keyboard, but their options cannot yet be chosen with the keyboard — a mouse or touch is currently required. Workaround: the same figures are available as text through the CSV downloads and the full data tables described above.
- Trend chart details. On the trend chart, the pop-up point values and the legend on/off toggles currently need a mouse. Workaround: the underlying figures are available through the CSV downloads and the full data tables.
- Keyboard focus indicator. On some controls, the outline that shows which element has keyboard focus can be hard to see against a dark background in dark mode.
- Theme buttons. The light/dark/system buttons work with the keyboard, but they do not yet announce to a screen reader which option is currently selected.
- A status label on the leaderboard. A small "fresh" status label on the leaderboard does not have enough contrast against its background.
- One control in dark mode. One control keeps a light background in dark mode, which leaves its label hard to read there.
What we have not tested
We have not tested everything, and we do not want to imply otherwise.
- Screen readers. We have not tested the site with a screen reader (such as VoiceOver, NVDA, or JAWS) in this evaluation, so we make no claim about that experience.
- Areas such as cognitive-accessibility guidance and some timing-related criteria were not assessed in depth.
This statement
This statement was prepared on August 17, 2026 and reflects testing of the site as described above. We will review it at least once a year, and sooner when the site changes in a way that affects accessibility.