Why Replit, Lovable and Bolt apps get rejected from the App Store

In March 2026 Apple quietly blocked updates for several "vibe coding" applications, among them Replit and Vibecode. The decision brought a basic problem into the open: apps generated by these tools are not real native apps, and Apple has no intention of letting them through.

What happened

On 18 March 2026, The Information reported that Apple had blocked Replit's updates on the App Store. Since its last update in January, the Replit app fell from first to third in the free developer tools chart. Vibecode, another vibe coding platform, was rejected outright.

Why Apple rejects these apps

Apple points at two specific rules in its guidelines:

The technical problem is plain. When a user types a prompt into Replit, the AI generates code that runs directly inside the app through an embedded webview. The user ends up with an app inside an app, one Apple never reviewed. That is precisely what guideline 2.5.2 forbids.

The disguised web app problem (guideline 4.2)

Beyond vibe coding, a wider problem affects every tool that generates web applications and then tries to publish them on the App Store. Platforms such as Lovable, Bolt and other no-code tools produce web applications wrapped in a native container, a webview.

Apple regularly rejects these apps under guideline 4.2, Design: Minimum Functionality. The rule requires apps to be useful and to offer a real app experience. A website in a webview does not clear that bar.

To stand any chance of passing review, a wrapped app has to include at least:

• At least two native features (push notifications, biometrics, in-app purchases and so on)
• Deep links that actually work
• An offline mode for the main screens
• A settings screen with a privacy policy and notification controls

Even with all of that, approval stays uncertain. And these apps do not hold up over time: Apple tightens its criteria regularly, and what passes today gets rejected tomorrow.

Is Apple being inconsistent?

Critics have pointed out that Apple allows code execution in its own apps: Swift Playgrounds, Pythonista, Safari with JavaScript, and the Shortcuts app. Apple treats those as sandboxed, well defined interpreters, as opposed to the much broader execution scope Replit offers. The distinction is real in technical terms, even if it can look arbitrary from the outside.

At the same time Apple built agentic coding support into its own IDE, Xcode, which says the company is not against AI in development. It wants the resulting apps to be genuinely native and to go through its review.

The answer is native development

The reality is simple. If you want an app that will be accepted on the App Store and stay accepted, you need a native app. Not a website wrapped in a webview. Not code generated and executed inside a container. Real Swift for iOS, real Kotlin for Android.

AI is still an excellent way to move faster, in the hands of a developer who knows the guidelines, the architectures and the conventions of each platform. That is the difference between an app that gets through review once by luck and an app that gets through on every update because it was built to the rules in the first place.