Accessibility audits for mobile apps
Evidence-based accessibility audits for mobile apps. Every finding carries the measurement that produced it, mapped to WCAG 2.2 and EN 301 549 — and every check we could not run is published as a gap, never as a pass.
$149.99 per audit · No account required · See the full output before you buy
The sample report — what one audit hands back:
A real finding — not a screenshot
Evidence
Illustration, not a screenshot: the app is fictional and this frame is generated from the same pixel rectangles the report publishes.
Reproduced in 3 of 3 attempts
Every finding a report can publish today comes from a deterministic rule that cannot hallucinate. The agent layer and the device confirmation are what the pipeline is built to do next, and they are marked as such below rather than described as if they had already run.
Your app built for the iOS Simulator and zipped — not an .ipa. The Mac app refuses the wrong file type in front of you.
no account · no repository access
Missing labels, labels that hide their visible text, controls with no role, unlabelled fields, targets under 44pt, reading order — measured, with the value found and the value required.
13 active rules · WCAG 2.2 · EN 301 549 · no model
Built to drive your app across tasks and environments so a failure arrives as a recording, not as a claim. None of the nine agent rules is active yet.
evidence policy: failure frame only
Built to reproduce critical findings on a physical device before they ship. What cannot be reached is already published as not testable, never as a pass.
not_testable is never silent
Every finding with its evidence, and two numbers that are never combined: a readiness score and a coverage percentage.
48h target · score withheld under 60% coverage
Then, on audit completed:
Deletion is triggered by the audit finishing, not by a retention clock. A run we could not complete holds your build for 72 hours so you can attach a fixed one to the same audit, then deletes it too. What stays is the measurement — it is linked to your audit, so it is pseudonymous rather than anonymous, and it is what lets a second audit tell you which findings you actually fixed. The full retention rules →
EAA · Europe
The European Accessibility Act has applied since June 2025. It reaches a listed set of services offered to EU consumers — e-commerce, consumer banking, e-books, telecoms, access to media, passenger transport and emergency communications — wherever the company itself is established. Companies under ten people that stay under the €2M ceiling are exempt as service providers. For everyone else on that list, penalties are set per country.
ADA · United States
There is no US regulation telling a private app what to meet — Title III has no web or app rule, and the circuits disagree about whether a service with no physical place behind it is reachable at all. What does exist is a filing pattern, an appellate decision that covers an app by name, and a Title II rule with dates in it. A recorded, reproducible audit is the artifact that survives a dispute, because it is evidence rather than an opinion.
$149.99
One audit — one app, one platform. Delivered within 48h.
$99.99/mo
2 audits/month + a regression diff on every build.
$299.99/mo
5 apps · white-label PDF export.
No. It is a technical readiness assessment — a measured, reproducible view of where your app stands against WCAG 2.2 and EN 301 549. It is not legal advice, and we never claim certification.
Those are excellent for a web DOM. They cannot install and drive your iOS app, turn on VoiceOver, or record a checkout that traps a screen-reader user. Closing that gap is what this product is built around. What ships today is the deterministic layer — twelve active rules measured against the accessibility tree your build exposes — and the report names the layer behind every row, so you always know which one you are reading.
One artifact: your app built for the iOS Simulator and zipped — what xcodebuild -sdk iphonesimulator produces, not an .ipa. The app shows you the command and refuses the wrong file type in front of you, not two days later. If you would rather run one thing than assemble two, our packaging script is public on GitHub as accessproof-package: about a hundred lines of bash, MIT, no network calls, meant to be read before it is run.
You attach a fixed one to the same audit — up to twice, within 72 hours, at no further cost. Being straight about the other half: the credit is spent when we accept the build, and a failed run does not return it, so what we give you is the attempts rather than a refund we have no mechanism to make. Failures on our side are re-run automatically and do not use up those attempts.
They run on isolated simulators and are deleted the moment the audit completes — the report stamps the exact UTC time. The visual evidence is embedded in the report itself, and we keep no copy of it. What we do keep is the measurement: the finding rows, the score and the coverage. That is what lets a later audit tell you which findings you fixed, and it is linked to your audit, so it is pseudonymous rather than anonymous. Only the benchmark projection is de-identified, and you can have either the benchmark inclusion or the whole record dropped by asking.
All questions, including the ones about what this cannot do →