Our own audit

• At most three things on screen.

  One idea per line, very large type. Aphasia-friendly formatting and Hick's law both point the same way: fewer choices, faster and surer decisions.

• No timer, ever.

  A clock on someone who cannot find a word adds pressure and makes the block worse.

• No transcript of their own speech.

  Seeing your own disfluent speech written out is discouraging, and it is not what the caller needs.

• Two real choices, never yes or no.

  Buttons 88 pixels tall, at most four words each, plus "Something else".

• Atkinson Hyperlegible for everything the caller reads.

  Designed by the Braille Institute to make letters hard to confuse.

• Stop is on screen whenever it speaks for the caller.

  Its own colour token, with contrast checked separately, because it is the one control that must always work.

Sources: Rose et al. 2003, 2011Aphasia Institute, SCAWCAG 2.2

Every text and state colour, in both themes, checked against WCAG 2.2. This runs before every build and the build fails on any violation. Last run 2026-09-29.

The amber “worked out” evidence colour is below 3:1 in light mode. It is never the only signal: every evidence mark also has an icon and a text label, and the colours were chosen with a colour-blindness validator after our first draft, green and red, failed it.

axe-core 4.13.0 (WCAG 2.2 AA and best-practice rules), run on 2026-09-27 against every page in both themes. Automated tools catch only some real problems, so this is a floor, not a verdict.

Everything above was run in Chromium, on Windows, at desktop and phone widths, in both themes. We have no Apple device, so Safari and iOS are untested, and rather than guess we have made sure the part a visitor most needs does not depend on anything exotic.

The practical consequence: a judge on an untested browser, or one who declines the microphone prompt, still sees three real calls with every event they produced. That path was made to need nothing more than an audio tag precisely because it is the one we cannot verify everywhere.

• No testing with people with aphasia, and no review by a speech-language pathologist. What we did instead is check each design decision against research done by people who worked with them, and say plainly what that does and does not establish: the parameters, against the literature.
• No testing on Safari or iOS, because we have no Apple device. See above for what depends on what.
• No full manual screen-reader walkthrough. Live regions announce the caller's state and the replay captions, but that is not the same as a real audit.
• Voice input only. A caller who cannot speak at all needs a different product.