Drafted with AI assistance and reviewed, fact-checked and edited by hand before publishing - see our AI transparency page for how we use AI across the app.
It's easy to make an accessibility checklist go green: add descriptions for images, make sure text is easy to read against its background, give every button a visible outline when you tab to it, run an automated checker, done. We do all of that - but we've come to think of it as the floor, not the job. A spelling app is used by children who find reading hard for very different reasons, and "technically ticks every box" doesn't mean a dyslexic seven-year-old, or a child who freezes under a countdown timer, can actually use it. Over the last few weeks we've shipped a run of accessibility features, and this post is about what we built and why we think the distinction between meeting a standard and being usable matters.
The floor: standards we don't compromise on
This part isn't glamorous, but it's the foundation everything else sits on:
- One focus outline, everywhere. When you move around the app using a keyboard or a switch device instead of a mouse or touch, the item you've landed on should be obviously highlighted. We used to have three different highlight colours scattered across the app depending on which screen you were on, and a couple of spots removed the highlight altogether without putting anything back - invisible unless you actually navigate that way, but a real barrier for someone who does. Every interactive element now uses the same clearly visible outline, with no exceptions.
- Text that's genuinely readable. Several of our greys for secondary text were only readable if you already knew what the words said - technically visible, not genuinely legible. We've replaced them with a darker shade that's comfortably readable at a glance.
- Motion you can turn off. Your phone or browser has a setting that says "reduce motion" for anyone who finds animation distracting or uncomfortable. We now honour that setting everywhere in one go - coin bursts, level-up flourishes, spinners - instead of the handful of spots we'd remembered to handle individually before.
- Checks that run automatically, every time. Every change to the app is automatically scanned for accessibility problems before it goes live, and a separate check stops new colours or styles sneaking in that don't match our approved, already-accessible palette.
None of that is novel. It's table stakes, and we treat it as non-negotiable - but it's also the part a computer can check for you, which is exactly why it isn't where the real work is.
Where the standard runs out: reading doesn't work the same way for every child
The official accessibility guidelines have nothing to say about which typeface helps a specific dyslexic child read more fluently, because there isn't one answer - it genuinely varies child to child. So instead of picking a font and calling it "the accessible mode," we built a choice: Standard, Clear (a typeface called Lexend, designed to reduce the visual crowding that trips up some readers), or OpenDyslexic (a font with heavier letter bottoms meant to stop letters flipping in the reader's eye). Alongside the font, a parent can set a colour overlay (cream, blue, rose, mint or grey - tinted backgrounds are a long-standing aid for some visual reading difficulties, and which one helps is individual) and letter and line spacing (normal, relaxed or roomy).
The part that matters more than any single option is that a parent can see it before committing to it - a live "how reading will look" preview updates instantly as they try each combination, using a real sentence rather than asking them to guess from a settings label. It's a per-child setting, synced across devices, so it travels with the child rather than living on one tablet.
Hearing a word isn't the same as hearing it clearly
Text-to-speech that only plays at one speed meets the letter of "provide audio alternatives" while missing plenty of children who need the word slower to catch it, or who need it broken into sounds rather than said as a whole. We added a speech speed control (Normal / Slow / Slower) that applies everywhere a word is spoken - hints, dictation, practice, the Skills Check - and a "Say each letter separately" Sounds button that spells a word out phonetically rather than just repeating it. Both are one tap away from a header button on the practice screen itself, not buried three menus deep in settings - because a control a child can't find in the moment they need it isn't really accessible.
The feature no accessibility guideline asked for: low-pressure mode
This is the one we think best makes the point. Nothing in the official standards requires removing a countdown timer from a game - a timer isn't automatically an accessibility problem. But for a child who freezes under time pressure, or whose processing speed doesn't match a generic timer tuned for the average player, a ticking clock can turn a game meant to build confidence into one more thing to fail at.
Low-pressure mode turns off the clock. Balloon Pop and Spell Buzz run untimed instead of racing a duration, the streak badges and leaderboard prompts that create their own quiet pressure disappear, and it's one toggle a parent sets per child - not a setting we make every family dig for. Getting it right meant checking every place in the app a game could surface a leaderboard link (we found 13, not the 3 we expected on the first pass), because a feature that hides pressure in 10 places and misses 3 doesn't actually remove the pressure - it just makes it inconsistent.
This is the difference we care about: a standard tells you what a screen reader announcement should say. It won't tell you that timed games are the wrong shape for some of the children using your app. That has to come from actually thinking about who's on the other side of the screen.
From "wrong" to "why": the recurring patterns report
Accuracy scores answer whether a child spelled a word correctly. They don't answer why they keep getting the same one wrong. The recurring patterns section of the progress report classifies repeated mistakes into transposition (letters swapped), omission (a letter dropped), insertion (an extra letter added), substitution, and doubled-letter errors, and surfaces the pattern next to the tricky-word ranking rather than leaving a parent to spot it by eye across dozens of attempts.
That distinction matters for how a parent or teacher actually helps: a child who consistently drops silent letters needs a different conversation than one who transposes letters under time pressure. A report that only says "got necessary wrong four times" doesn't get you there. One that says "this child keeps dropping doubled letters" does.
Controls that don't require finding the settings menu first
We also added a device-level text size toggle (A / A+ / A++) that sits in the header of the public home page, reachable before anyone has signed in or chosen a child profile at all. Accessibility settings that only exist after a login flow help the people who already know to look for them. A control any visitor can tap the moment the page loads helps everyone else.
Why we're writing this down
None of the individual pieces above were required by an official checklist. A dyslexia font choice with a live preview, an untimed mode for games, a report that names the shape of a mistake instead of just counting it - these came from asking "would this actually help the child in front of us," not "does this tick the required box." Standards are a good, necessary floor - they catch the failures that are obvious once you go looking (missing descriptions, hard-to-read text, a highlight you can't see). But the ceiling is set by paying attention to how differently children actually read, listen, write and cope with pressure, and building for that difference rather than for an average user who doesn't exist.
We'll keep doing both: holding the line on the measurable stuff (our automatic checks against rogue colours and styles only get stricter over time), and shipping the features that don't show up on any checklist but are the reason a child keeps coming back to practice instead of giving up on it.
If there's a way your child learns, listens, or copes with pressure that we haven't accounted for yet, tell us - every feature on this list started as feedback from a real family, not a spec.