How Apple’s App Store Crackdown on AI Is Quietly Reshaping Which Apps Survive

There comes a time, perhaps the third or fourth time a business modifies its justification for an action, when you completely lose faith in the initial explanation. For anyone watching Apple’s management of Replit, the AI-powered coding platform that hasn’t been able to release an iOS app update since early this year, that moment came quietly.

According to Apple, it has to do with safety. According to Apple, it has to do with uniform enforcement. However, the narrative continues to change, and in the meantime, Replit has dropped from the top spot in the App Store for developer tools to number four, losing actual revenue, real users, and a window of opportunity.

This is not a small software disagreement. It’s more akin to a policy identity crisis occurring within one of the world’s most valuable corporations.

Eighteen months ago, vibe coding—the abbreviation for AI-assisted app development, in which a user describes what they want and a tool like Replit builds it—was hardly a thing. The businesses in this industry are now worth billions of dollars, and the number of new apps submitted to the App Store increased by 60% year over year to over 550,000 last year—the most in ten years.

First-time builders who had never written a line of code before made up a sizable portion of that wave. People like Ruth Heasman, a graphic designer in the UK who couldn’t code and couldn’t afford to hire someone who could, so her app ideas lived only in her head for twenty years. That was altered for her by Replit.

A ghost-hunting augmented reality iOS game is one of her approximately twelve published projects to date. She takes pride in it. Additionally, since she doesn’t own a Mac, she wouldn’t be building at all if Apple’s recommended method—build using Xcode, submit for review, and distribute through the store—were the only option.

Apple’s argument begins to crumble at that gap between the user Apple envisions and the user who actually shows up. Only a few weeks after blocking Replit’s update, the company integrated AI tools from Anthropic and OpenAI directly into Xcode in February.

This might have been a coincidence. Maybe it wasn’t. From the outside, it looks more like Apple shielding its own development environment from competition while citing safety as justification than it does like principled enforcement.

When examined more closely, the safety argument itself falls short. A Replit user’s creations are not immediately installed on a phone. Every time a user taps an external link, it operates inside the app using the same web technology that Facebook and X have been using for years. Those apps have never been reported by Apple for displaying unapproved web content. To put it simply, the distinction Apple is making seems to be based on convenience.

Apple has previously engaged in and frequently prevailed in similar conflicts. In court, it outlasted Epic Games. For as long as it was permitted by law, it defied the EU’s sideloading directives. WeChat’s mini-app structure caused it to clash with Tencent. In each instance, it involved protecting the App Store from businesses attempting to circumvent it.

Users only need to open a laptop browser and use Replit, so vibe coding doesn’t need to figure out a workaround. In this instance, the wall Apple is defending doesn’t really keep water out. It simply reduces the iPhone’s usefulness as a workspace.

It’s difficult to overlook the larger economic logic that underlies all of this. With gross margins above 75%, the App Store is the hub of a services business that brought in $109 billion last fiscal year—nearly twice as much as Apple makes from selling its hardware.

Every app that is developed for the open web and made available without going through Apple’s review process is money that Apple never receives. Simultaneously, vibe coding is filling the App Store and creating its replacement. Anyone who is paying attention is aware of that tension.

The recently revised App Review Guideline 5.1.2(i), which expressly mandates that apps disclose and obtain explicit user permission before sharing personal data with third-party AI, is especially noteworthy. On the surface, the update makes sense: data privacy is important, and AI firms that handle user data should be held responsible.

However, the timing and the specificity—naming AI companies by name in a manner that the previous rule never did—indicate that Apple is tightening its hold just as AI tools begin to truly benefit the general public.

As this develops, it’s difficult to avoid thinking back to the 1990s, when Microsoft allowed the PC ecosystem to expand in all directions while Apple restricted its hardware. That sprawl appeared disorganized. Additionally, it taught a whole generation how to construct things. When Jobs returned, he rebuilt a business that gave people access to tools. The idea behind the iPhone, the App Store, and even the first Mac was to reduce rather than increase barriers.

It’s still unclear if Apple’s current strategy is the result of a well-thought-out plan or just the friction of a big company where various teams—App Store review, developer tools, policy, and legal—are pulling in different directions without anyone at the top pressuring them to align. According to Replit, Apple has changed its stated justifications for the hold several times since January, bringing up fresh concerns each time an earlier one was resolved.

According to Apple, its review team has been in constant contact with Replit and has spoken with him over the phone three times in the last two months. It is possible for both to be true. There is a difference between consistent reasoning and consistent communication.

People who use vibe coding tools to build will continue to do so. That much appears to be certain. The only real question is whether Apple will continue to develop the iPhone or if it will subtly distance itself from the most innovative period in software development since the App Store’s initial launch.