The screen looked great. Custom canvas buttons, tight typography, a neat little swipe to confirm. Then someone on the team turned on a screen reader and tried to buy something. The reader announced a blank region. The confirm control had no label. The font refused to scale because we'd pinned it in pixels. There was, functionally, no checkout.

We fixed it in a week, but the lesson stuck: mini-programs are unusually easy to make inaccessible, because so much of the UI is custom-drawn rather than made of native elements.

Why this hits harder here

A web page gets a lot of accessibility for free — semantic elements, focus order, text that reflows. A mini-program often doesn't. Teams rebuild buttons, tabs and modals to match the host's visual style, use canvas for performance, and hard-code sizes to keep a design pixel-perfect. Every one of those choices quietly removes something assistive tech was relying on.

And the users who pay for it aren't a small edge case. In many of the markets where super apps are the default way to reach services, people are on older devices, smaller screens, enlarged system fonts, and unreliable connections. Add an ageing population and low-vision users, and "accessible" stops being a compliance checkbox and starts being a conversion question.

ProblemWho it hitsWhat actually fixes it
Font pinned in pxLow vision, enlarged system fontRespect host font scale, use relative units
Canvas-only controlsScreen reader usersA real labelled element behind or beside it
Tap targets under 44pxMotor impairment, one-handed use on a busBigger hit areas, more spacing
Low contrast textBright sunlight, older screensCheck ratios against the actual brand palette
Icon-only buttonsEveryone, the first timeAccessible labels, not just tooltips
Gesture-only confirmationScreen reader, switch accessA tappable alternative that does the same thing
Silent loading statesSlow connections, cognitive loadAnnounce progress, don't leave a blank pause

The honest tradeoff

We won't pretend this is free. Matching a host's visual language pixel-for-pixel is a real design goal, and accessible labels, larger hit areas and reflowable text all push against it. Custom canvas is sometimes the only way to hit a frame budget in a mini-game. There are genuine tensions, and pretending otherwise is how teams quietly drop the work.

But most of the cost is front-loaded. Put labelled components, contrast-checked colors and scalable type into the component library once, and every new screen inherits it. Do it as a pre-launch audit instead, and you'll be fixing the same six issues on every release, usually under time pressure, usually incompletely.

We'd also argue against treating WCAG as the ceiling. It's a floor, and a useful one — contrast, focus, names and roles are measurable. The thing that actually tells you whether someone can pay is turning on the screen reader and trying, with the system font at 200%, on a cheap phone over a weak connection. That test takes ten minutes. It found more than any checklist we'd written.

Make it a default, not a phase

Start with the flow that matters most — checkout, booking, submitting a form — and walk it with assistive tech on. Fix the labels and hit areas there first. Then move the fixes into shared components so they can't be forgotten.

There's a platform angle too. Hosts differ in what they expose: system font scale, screen reader APIs, reduced-motion settings. A runtime that targets several hosts can normalize those into one set of accessible primitives, so the same mini-program stays usable whether it lands in a messaging app or a payments app. Accessibility shouldn't be the thing you re-solve in every market.

Sources