Apple said no once.Here's what I changed before asking again

What actually happened

Build 11 went to App Review and came back rejected. Two findings, both specific and both fair. This is the honest walk-through of what Apple flagged and what I did about it, because I'd rather be frank about a rejection than pretend the road to the App Store was a straight line. It wasn't, and it shouldn't have to look like one.

Finding one: how we tell you about third-party AI

Refuel AI, the conversational feature in the app, is powered by Claude, a large language model from Anthropic. Apple's guidelines (sections 5.1.1(i) and 5.1.2(i)) require explicit, informed consent before sending any user data to a third party like that — and the review found our existing disclosure wasn't explicit or upfront enough.

They were right to flag it. The fix is a new screen, RefuelAIConsentGate, that appears before your very first Refuel AI request and:

• Names the actual provider — Anthropic, and the Claude models behind Refuel AI — rather than a vague "our AI partner."
• States plainly what gets sent (the message you type, and safe, coarse context like a check-in trend) and, just as clearly, what never gets sent (your name, your identity, anything from HealthKit).
• Links directly to the full privacy policy for anyone who wants the complete picture.
• Requires an active tap on "Allow and continue" — there's no default-yes, no dark pattern, no pre-checked box.

I also added a second, defense-in-depth check inside the code path that actually sends a request, so even if the consent screen were somehow bypassed, the request still can't go out without recorded consent. And I updated the web privacy policy, in both English and Ukrainian, to name Anthropic explicitly and describe what "equivalent protection" means for data sent to them.

Finding two: HealthKit did nothing on iPad

The second finding (guideline 2.1(a)) was more mundane but just as real: tapping "Connect" for HealthKit on iPad did nothing at all. No system permission sheet, no error, nothing. The cause was almost embarrassingly small — the app's entitlements file was missing the actual com.apple.developer.healthkit entitlement, so iOS silently refused to even show the authorization prompt. I added the entitlement, and on top of that added a real fallback: if access still isn't granted for some other reason, you now get a clear alert with a direct route to Settings, instead of a button that quietly does nothing.

Why I'm writing this up instead of just shipping it quietly

Because the point of the consent gate is exactly the point of this post: telling you plainly what's happening rather than assuming you won't notice or won't ask. Refuel AI already never sees your name or your HealthKit data, and it was never going to. What was missing was making that as obvious to you, in the app, as it now is to Apple's reviewers. This fix makes the app more honest, not just more approvable.

Why the version jumped to 2.0.0

The resubmission carries a version bump from 1.0.0 to 2.0.0, and the build number from 10 to 12, across the main app, the widget and the Watch app. That's purely bookkeeping, not a claim about features: builds 10 and 11 were already consumed in App Store Connect under version 1.0.0, so this resubmission needed a fresh version train. I verified both a debug simulator build and a signed release device archive carry 2.0.0 (12), and that the archived app is signed with the HealthKit entitlement actually present this time.

Where and when

• Sept 7, ~13:56 — AI consent gate (RefuelAIConsentGate) and HealthKit entitlement fix implemented and verified in the simulator; 246 unit tests still passing (92eb15a).
• Sept 7, ~13:57 — Merged to the release branch (PR #19).
• Sept 7, ~14:17 — Version bumped to 2.0.0, build 12, across app, widget and Watch targets, ahead of resubmission (e1f9a76).

What this means if you're using or about to use the app

The very first time you try Refuel AI, you'll see the consent screen and have to actively agree before anything is sent anywhere. If you connect Apple Health on an iPad, the system prompt now actually appears. Nothing about what data Refuel AI does or doesn't see has changed — only how clearly, and how explicitly, you're told about it before it happens.