App releases are up 104% across the App Store and Play Store compared to a year ago. On iOS alone, it's 89%. Productivity apps have moved into the top five categories for the first time. TechCrunch's working hypothesis, and mine, is the same one: AI tools have crossed a usability line, and the people who never shipped before are shipping now.
That part is great. The part nobody seems to be writing about is what happens after xcodebuild finishes.
The boom is real, the bottleneck moved#
A year ago the limiting factor for a one-person app was can you build the thing. Today, with Claude Code and Cursor and a weekend, you can. The output is a real iOS or Android binary, often signed and on a TestFlight invite, and often shockingly close to functional.
What's not in the box: a privacy manifest that matches the SDKs you ended up importing. Five locale-correct screenshots at the right pixel size. A description that doesn't trip a metadata reviewer. An icon that's actually 1024×1024 with no alpha channel. An age rating that matches the content. A build that doesn't import a library Apple flagged six months ago for using a banned API.
None of those steps got faster because the code got faster. The reviewer queues are the same humans (and the same review automation) they were in 2024. And every one of those queues is now seeing roughly twice the volume.
What indie devs are actually hitting#
I run a submission tooling company, so I see the rejection patterns. Two are dominating right now.
Privacy manifests that don't match imports. The dependency graph from a vibe-coded project looks nothing like the manifest scaffolded by Xcode or stitched together by an AI assistant. Apps are getting bounced for declaring SDKs they don't use, omitting ones they do, or quietly importing a "required reason API" without the corresponding declaration.
Screenshots and metadata that look AI-generated. Apple has gotten visibly stricter on listings that read like copy-pasted GPT output — keyword-stuffed descriptions, vague feature lists, screenshots that don't show the actual app. None of these are new rejections, but the volume is.
The pattern isn't "AI builds bad apps." The apps are often fine. The pattern is "the people building with AI haven't internalised everything that happens between it works on my device and it's live." There's a lot of that work, and it's exactly the part you can't really skip.
The stora angle#
This is roughly why Stora exists. The compliance engine reads your bundle, your manifest, your imports, and your listing draft together — and flags the divergences before App Store Connect or Play Console does. The screenshot pipeline generates locale-correct screenshots from your build, not stock images. The store listing model writes copy that sounds like a real person describing a real app, not a marketing keyword salad.
If you're shipping ten apps this year because the AI made it possible, the submission part still scales linearly with your patience. We're trying to fix that.
Closing#
The good news: the AI tooling shift is genuinely lowering the floor on who can build a mobile app. The honest news: the App Store and Play Store didn't get the memo. The bottleneck has moved from can I build this to can I push this through review without losing a week. Plan the second half of that.