Project Log #14: I Scored 30 Apps on Accessibility. Here's the Public List.

#ai #agents #webdev #automation

Day 14. The audit doubled in size. The accessibility score list is now public on GitHub.

Last week, I published my first accessibility audit. 15 apps. A simple A-to-F score. The response surprised me.

Developers reached out asking how their apps scored. A few wanted to know how to improve. One person told me they'd never thought about accessibility labels before reading my build log.

So I spent the next few days doubling the audit. 30 apps. Same methodology. Same scoring system. The results are now public on GitHub.

The Expanded Audit

I added 15 more apps across new categories. Health apps. Education apps. Government services. More local apps.

Category	Best Apps (A/B)	Worst Apps (D/F)
Messaging	WhatsApp, Telegram, Google Messages	—
Email	Gmail, Outlook	—
Productivity	Slack, Notion, Google Calendar	—
Navigation	Google Maps, Waze	—
Banking	—	5 out of 5 tested scored F
Health	—	3 out of 4 scored D or F
Government	—	4 out of 4 scored F
Food Delivery	Uber Eats, Deliveroo	Local apps scored D
Education	Duolingo, Khan Academy	Local education apps scored D

The pattern held. Global apps score high. Local apps score low. Every single banking app tested scored an F.

The Public List

I've published the full accessibility score list on GitHub. It's in the README under a new section called "Accessibility Audit."

Each app gets:

A letter grade (A through F)
A short explanation of why it scored that way
Whether the UI tree is navigable by screen readers
Whether my agent can automate it

This list is now the most detailed public resource I know of that connects AI agent capability to accessibility compliance.

What This Means for Developers

If you're building a mobile app, here's what you should know:

Accessibility labels are not optional. They're not a "nice to have." They determine whether millions of visually impaired users can use your app—and increasingly, whether AI agents can interact with it.
Adding labels is not hard. A single content-desc attribute on a button takes seconds to add. The reason most apps score F is not technical difficulty. It's that nobody prioritised it.
The gap is a business opportunity. An app that scores A is usable by more humans AND more machines. As AI agents become more common, accessible apps will have an automation advantage.

What's Next (Day 15)

Start reaching out to developers of F-scoring apps with specific, actionable feedback
Add a "How to Improve Your Score" guide to the repo
Begin testing whether my agent can now handle multi-app workflows (e.g., "Copy my bank balance from the banking app and send it to Mom on WhatsApp")

The Repo

👉 github.com/Dexter2344/phone-agent

The accessibility audit is now a permanent section of the README. 30 apps scored. Public. Free. For anyone to use.

This is Day 14. The list is out. The conversation is starting.