EZCare is a wellness app with an AI companion in it. People check in on their sleep, energy and mood, follow a daily plan, and talk to EZBuddy when they want guidance. It runs on iOS, Android and the web from one codebase.
The build was solid. Getting it past App Store review was its own piece of work, and each thing Apple flagged taught us something worth writing down.
1. Words matter more than features
The first issue was health wording. Some of our copy read like the app could tell you something medical about yourself. It couldn't, and it was never meant to, but the reviewer reads the words, not your intentions.
We rewrote the copy around wellness, habits and routines, and added clear disclaimers where the AI gives guidance. Nothing in the app changed except the language, and that was enough.
2. Purchases have to line up exactly
The second was an in-app purchase ID that didn't match what was set up in App Store Connect. On a test device it's easy to miss, because everything else around the paywall works.
We corrected the setup, then went further and rebuilt the purchase check around the actual transaction, so the app unlocks based on what the store confirms, on both iOS and Android.
3. Keep your AI models current
The third surprised us: the app still referenced an AI model that had been retired. Reviewers test the features they see, and a companion that fails to answer looks broken.
We moved to current models and, while we were there, put every AI call behind a server function. The key never sits on the phone, each user has a rate limit, and usage is logged.
What we do differently now
Before we submit any app with AI or health features, we read the copy the way a reviewer would, check every purchase ID against the store, and confirm the models we call are still supported. It takes an hour and saves a week of back and forth.
