Beyond AA

We publish no video, and promising a British Sign Language interpreter for every future video is a commitment that never gets cheap. We caption everything instead, and we would revisit this if an NHS contract asked for it by name.

We publish no video. If we ever do, our production rule is to write the narration so it already says what the screen shows, which is the cheaper and more reliable answer to the same problem.

We have no video today, but we are adopting this ahead of time: any future video gets a written transcript from day one, because the script already exists and a transcript costs almost nothing on top of it.

Nothing on our site streams live audio, and nothing is planned. A page nobody maintains reads as coverage while holding nothing, so we are not adding one.

The technique this criterion asks for depends on a machine-readable format that no browser or assistive technology actually reads today, so claiming it would not help anyone using our product.

We are raising our body text to the stronger contrast bar where it is a simple colour step, but not across every coloured panel: that would undo contrast decisions we already tuned carefully, including a signed-off welcome colour, for a gain that is mostly theoretical there.

We have no audio today, and we are adopting the rule now: no music behind the voice in any future video, because a music bed makes speech harder to follow for people with auditory processing differences, who are common among our users.

Rather than build our own colour picker, we would rather make the product properly respect the contrast and colour settings you have already set on your own device.

Our website and our printed report already keep lines of text short and easy to follow. The gap is inside the app, where some cards and the report screen still run wider than we would like.

We keep text ragged rather than justified everywhere we can, on screen and in the printed report, because justified text creates uneven gaps between words that are harder for dyslexic readers to follow, and dyslexic parents are exactly who this criterion is for.

The space between paragraphs used to be too tight across our website, our printed report and the app. We widened it on all three in August 2026, setting it against each surface's own line height rather than one number everywhere. It is partly met rather than met because only the website version is measured automatically, and that check is one we watch rather than one that can block a change.

Zooming in to the 200 percent equivalent causes no page to scroll sideways. In August 2026 we made that an automatic check rather than a one-off measurement, and included the report page and a question page for the first time. It stays partly met because the check is one we watch rather than one that can block a change.

Two articles used to show a whole information table as a picture. Both are now real text you can select, resize and have read aloud, and a test keeps them that way. It stays partly met because that test reads image file names and cannot see words inside a picture with an ordinary name.

Every control in our product already works from a keyboard alone, with nothing that needs a drag or a mouse-only gesture, so meeting this in full costs us nothing extra, and we now block any change that would reintroduce one.

The app itself has no time limit anywhere. Our website's waitlist form has a hidden minimum submit time that can silently drop a genuine, fast signup behind a success message, and we are fixing that.

We already avoid pop-ups and anything that interrupts you unasked. The one gap is a demo animation in the app's welcome tour that loops with no pause control, which is worth a small fix rather than a whole project.

If your sign-in link expires, we already send you back to exactly where you were with your saved answers untouched. What we have not confirmed is what happens to anything you were mid-way through typing at that exact moment.

Your answers are saved as you go, one question at a time, and your session lasts well over a day even if you walk away, and we now test that the session length can never quietly shrink.

Nothing on our site or in our app flashes fast enough to be a seizure risk, a real concern for children more than adults, and we check every animation stays well under that speed.

Both the app and this website follow your device's reduce-motion setting through one shared rule, and the app's own Reduce motion setting does the same. The two animations driven by script, which that rule cannot reach, the report's tab underline and the chat's scroll, now check both settings themselves. It stays partly met because no test yet fails when a new animation ignores the setting.

Most of our product already tells you where you are, with breadcrumbs and a progress bar. The app's own menu does not yet mark your current page, and children's profile tabs need a clearer label than colour alone.

Our printed report already names every link clearly on its own. On screen, a repeated 'Read the full definition' link does not say which term it is about, and we are fixing that.

A parent or a school reading the report by its headings hits a few tabs where we deliberately hid the section heading. We are putting those headings back.

We are fixing the underlying problem, a fixed header or footer sometimes covering the very thing you have just tabbed to, because it is worth doing on its own. We will only claim the stricter, enhanced version once something actually measures that nothing is ever covered.

Our default keyboard focus ring is the right size and colour in both apps, we removed the handful of thinner overrides that fell short of it, and a check blocks that thin pattern from coming back. It stays partly met because the ring as actually drawn is measured on only some screens, by a check that reports rather than blocks.

The tap targets on our 75-question profiler are smaller than we would like today. Getting this right matters more to us than to most products, because motor skills are one of the very things our profiler assesses.

Nothing in our product works with touch only, or with a mouse only, and we block any change that would restrict how you interact with it.

This is the criterion our whole product exists to satisfy: a plain-English glossary for clinical and school language, tapped open wherever it appears. The mechanism is built and protected from regressing; we have barely started filling it with real terms rather than our own product words.

Initials like ADHD appear throughout our content with nothing yet spelling them out on first use. Our glossary already has a field ready to hold the full name; nothing has been written into it yet.

We score wording with the Flesch-Kincaid formula, converted to a UK reading age. Inside the app, wording is held at a reading age of 11, the NHS target, and any piece that reads older fails the app's type-check, one of the checks a change must pass to be merged. The NHS asks for 9 to 11 and accepts 11 to 14 where information is hard to make simpler, so consent and privacy wording is held at 14 the same way, because shortening it can drop part of what we are promising. A short named list of exceptions, all of it consent and privacy wording, is each held at its own ceiling so it cannot quietly get harder. Long-form articles on the public site are written for parents researching a topic, so their limit is 19, held the same way. The AI chat and the written report are told to write for a reading age of 12, and the report's overview for 9, but nothing yet measures what they actually write. PARTIAL is the honest status because the checks read the app's wording one piece at a time, not the finished pages as a parent sees them, and not the report, which is written fresh for each child. We are not promising a separate easy-read version of the whole site, which we could not keep maintained.

We have not found any word in our content whose pronunciation is genuinely ambiguous, so there is nothing yet for this criterion to apply to. We have not done the content review that would settle it either way.

The profiler moves you to the next question by itself, which is what makes a long questionnaire manageable. A "Keep moving on" switch on every question turns that off, and once you turn it off it stays off. With a keyboard, nothing moves on until you press Enter. A test fails if the switch stops working or stops remembering your choice.

Where help exists, it sits in the same place every time. A handful of screens, including the safety questions, offer no help at all today, and that is the gap we are closing.

Help is genuinely good where we have built it: real guidance first, before anything automated. It is missing on exactly the most frightening steps, the screener and the safety questions, which is where it matters most.

You can go back and change an answer during a grace window before it is final, and you can delete a child record yourself, including one you added twice by mistake. Two slips still lose what you typed: removing a comment has no undo, and leaving a page before you save loses a draft.

You sign in with a link we email you, or, if you would rather type something, a short code we email instead. Neither asks you to remember or work anything out: the code is one field you can paste into or let your device fill, and there is still no password and no CAPTCHA anywhere in the product. We test that none of those can quietly sneak back in.