Accessibility

What we have built

Each feature says how we check it and where it falls short.

Reading and understanding

A reading age target for each audience

We aim for plain English everywhere. App screens aim for a reading age of 11, the NHS target, and consent and privacy wording for 14. On the website, most pages aim for 11. Pages for professionals, programme pages, the about page and legal pages aim for 14, and longer articles for 19.
Measured by one formula, with no AI

We score wording with the Flesch-Kincaid formula, turned into a UK reading age. A test fails if the app's messages or the website's articles go over their limit. Other website pages and the report's fixed wording are measured but not held to a limit. We also score each area of the app as a whole.
The AI is told, then measured

The chat and the report are told to write for a reading age of 12, and the report's opening summary for 9. We measure and store the reading age of each report the AI writes. If the measuring fails, the report still goes out, with no score. A report that reads too hard is not stopped. We check the chat's replies in test runs, not live.
A glossary where the words are

Harder words in your child's profile and report open in place to show a plain meaning. The glossary holds over 250 terms, and a person has not yet checked most of them. Short forms such as ADHD are not spelled out yet where they appear.

Your own display settings

In the app you can choose light or dark mode, bigger text, less motion and less see-through layering. The website offers the same, but its theme follows your device. There is no high-contrast option yet.
Windows contrast themes

Focus outlines and your chosen answers stay visible in forced colour themes, such as Windows high contrast. Our tests for this cannot stop a change going in. For the app, a failure stops that change reaching the live app unless we override it and record why.
Printing comes out light

Pages print as dark text on light paper, even in dark mode. A test fails if that stops working.

No passwords, no puzzles

You sign in to the app with a link or a code sent to your email: no password, and no puzzle (CAPTCHA). A test fails if either comes back.
Errors that say where they are

A form's error message is linked to its field, so a screen reader reads them together. A test checks this on the app and the website. The app's messages do not always say how to fix the problem yet.
Your browser can fill it in

Name, email and organisation fields carry standard hints, so your browser can fill them in. A test fails if one goes missing.
The questionnaire works without JavaScript

In the app you can answer every question, add a child and make the sharing consent choice with JavaScript off. A test checks each kind of question this way before any change can go in. Browser tests check adding a child, comments and the sharing choice once a change is in. A failure stops that change reaching the live app unless we override it and record why. The website's early access sign-up and enquiry forms work without JavaScript too.

How we build pages that work without JavaScript

Focus follows you

When the next question appears, your keyboard and screen reader move to it. A test checks this in the questionnaire. We have not checked the rest of the app this way yet.
You choose the pace

Questions move on by themselves once you answer. A Keep moving on switch turns this off. It appears when you go back to a question, and once off it stays off. With a keyboard, nothing moves on until you press Enter.
Nothing flashes, any input works

Nothing flashes, and touch, mouse and keyboard all reach the same controls. A test fails on the common ways to break either of these. It cannot see animation driven by script, or the frame around each question.

Shared components, on their own

Most shared building blocks have a demo page, and an automated accessibility checker tests each one, contrast included. A failure blocks the change unless an admin overrides it. Parts built only inside the app, and the frame around each question, are not tested this way.
Reading order, written down

Each app screen has a written record of the order a screen reader meets its content. An AI agent has read them all once, but no person has listened to them yet.