Something strange started happening to Apple’s App Store in early 2025. The volume of apps submitted for review surged — not gradually, but in a sharp upward spike that caught the attention of developers, analysts, and Apple itself. The culprit wasn’t a new iPhone launch or a hot platform trend in the traditional sense. It was vibe coding.
The term, coined by former Tesla AI director Andrej Karpathy in February 2025, describes a method of software development where a human describes what they want in plain language and an AI model writes the actual code. The human “vibes” with the output — accepting, rejecting, or tweaking results without necessarily understanding every line. It’s programming by intention rather than expertise. And it has opened the floodgates.
According to a report from AppleInsider, App Store review submissions climbed significantly throughout 2025, with Apple’s review team facing an unprecedented wave of new applications. The increase has been attributed in large part to the proliferation of AI coding tools — including GitHub Copilot, Cursor, Replit, and Anthropic’s Claude — that have made it possible for people with minimal programming backgrounds to build and submit functional iOS applications. What once required months of learning Swift or Objective-C can now, at least superficially, be accomplished in an afternoon.
The numbers tell a striking story. Apple processes millions of app submissions annually, but internal estimates and third-party analyses suggest that 2025 saw a year-over-year increase that far outpaced anything driven by previous platform shifts. Some estimates place the growth in new submissions at 30% or more compared to 2024. Not all of these apps are making it through review — far from it — but the sheer volume has strained a system that was already one of the most demanding quality gates in consumer technology.
For Apple, this creates a genuine operational problem. The company’s App Store review process combines automated scanning with human evaluation. Reviewers check for compliance with Apple’s guidelines, which cover everything from privacy requirements and content standards to technical performance and design quality. More submissions mean more work. And not just more work — more low-quality work that still requires human eyes to evaluate and reject.
The quality issue is where this story gets uncomfortable for AI optimists. Vibe coding is extraordinarily effective at producing something that looks like a working app. It is considerably less effective at producing something that is a good app. Many of the AI-generated submissions arriving at Apple’s door suffer from familiar problems: cloned functionality from existing apps, shallow user experiences, poor error handling, and security vulnerabilities that a seasoned developer would catch but a novice guided by AI would not. As AppleInsider noted, the trend has forced Apple to grapple with how to maintain its curation standards while processing a dramatically larger pipeline of incoming software.
Apple hasn’t publicly disclosed specific rejection rate changes tied to vibe-coded apps. But developers who frequent forums and communities like the Apple Developer Forums and indie dev groups on social media have reported noticeably longer review times in recent months. Some have speculated that Apple has begun deploying additional automated screening tools specifically designed to flag characteristics common in AI-generated submissions — repetitive code patterns, boilerplate privacy policies, and UI layouts that closely mirror tutorial templates commonly used by AI coding assistants.
The broader context matters here. Vibe coding didn’t emerge in a vacuum. It arrived at a moment when the major AI labs were locked in an intense competition to prove the practical utility of their large language models. OpenAI, Google, Anthropic, and others have all positioned code generation as a flagship capability. GitHub Copilot, powered by OpenAI’s models, crossed 1.8 million paying subscribers in 2024. Replit, which targets less experienced developers, has leaned heavily into AI-first workflows. Cursor, a newer entrant, has built its entire product around the idea that AI should write most of the code while the human steers.
These tools are genuinely impressive. They can scaffold entire applications from a text prompt, generate working API integrations, and produce UI code that compiles and runs on the first try. For experienced developers, they’re powerful accelerants. For inexperienced ones, they’re something more complicated — a way to produce output that exceeds their ability to evaluate it.
That gap between production capability and evaluation capability is the core tension driving the App Store flood. A person who can’t read Swift can now generate a Swift application. But they can’t audit it for memory leaks, assess whether it handles edge cases gracefully, or determine if it inadvertently collects user data in ways that violate Apple’s privacy guidelines. The AI doesn’t reliably catch these issues either. So the burden shifts downstream — to Apple’s review team, and ultimately to users if flawed apps slip through.
Some in the developer community have responded with frustration. Established indie developers, many of whom have spent years building reputations and refining their craft, see the influx as a threat to discoverability. When the App Store is flooded with low-effort clones and thin wrappers around AI-generated code, legitimate apps get buried. The signal-to-noise ratio degrades. And Apple’s search algorithms, which already favor larger publishers and apps with existing download momentum, become even less friendly to small creators trying to stand out.
“It’s like the SEO spam problem, but for apps,” one indie developer posted on X in late March 2025. The analogy is apt. Just as AI-generated content has overwhelmed Google’s search results with low-quality pages optimized for ranking rather than reader value, AI-generated apps threaten to overwhelm the App Store with software optimized for submission rather than user satisfaction.
Apple has tools to fight back. Its review guidelines are detailed and actively enforced. Guideline 4.3, which addresses spam and app cloning, has long been used to reject submissions that duplicate existing functionality without adding meaningful value. Apple could tighten enforcement of this guideline — and there are signs it already has. Developers have reported an uptick in 4.3 rejections in early 2025, though Apple hasn’t confirmed whether this reflects a deliberate policy shift or simply the natural result of more clone-like apps hitting the queue.
There’s also the question of whether Apple will require developers to disclose AI involvement in their submissions. No such requirement exists today, but the idea has been floated in developer circles and by some industry analysts. A disclosure mandate could help Apple’s reviewers prioritize their attention and apply appropriate scrutiny. It could also, critics argue, create a stigma around AI-assisted development that would be counterproductive — penalizing legitimate uses of AI tools alongside the low-effort spam.
Google faces a parallel challenge with the Play Store, though its historically lighter-touch review process means the flood dynamics play out differently. Google has relied more heavily on automated scanning and post-publication enforcement, pulling apps after they’ve already been listed if problems are detected. This approach may prove less sustainable as the volume of AI-generated submissions grows across both platforms.
The financial incentives fueling the vibe coding boom are straightforward. Building an iOS app and listing it on the App Store has never been cheaper. The cost of AI coding tools ranges from free to roughly $20 per month. An Apple Developer Program membership costs $99 per year. For someone hoping to generate even modest ad revenue or in-app purchase income, the barrier to entry is now almost negligibly low. This is, in one sense, a democratization of software development — a genuinely positive expansion of who can create and distribute software. In another sense, it’s an invitation to a gold rush mentality where quantity overwhelms quality.
Not everyone in the industry sees this as a crisis. Some argue that the flood is temporary — that Apple’s review process will adapt, that the worst vibe-coded apps will be filtered out, and that the net effect will be a larger and more diverse app marketplace. Karpathy himself has been broadly optimistic about vibe coding’s potential, framing it as a natural evolution in how humans interact with computers. Build things by describing them. Let the machine handle the syntax.
But the optimistic case depends on gatekeepers doing their jobs effectively. And right now, the primary gatekeeper — Apple — is being asked to do significantly more work with what appears to be roughly the same resources. Apple’s App Store review team is large but not infinitely scalable. Every hour spent evaluating a low-quality vibe-coded submission is an hour not spent on a legitimate app from a serious developer. The opportunity cost is real.
There are also deeper questions about what happens to the craft of software development if vibe coding becomes the dominant mode of app creation. Senior engineers and computer science educators have raised concerns that a generation of developers may learn to prompt AI models without ever understanding the underlying principles of the code being generated. This matters less for simple apps — a to-do list or a weather widget — but matters enormously for applications that handle sensitive data, manage financial transactions, or control hardware.
The security implications are particularly worrying. AI coding assistants have been shown in multiple studies to generate code with known vulnerability patterns. A 2024 study from Stanford researchers found that developers using AI assistants produced less secure code than those working without them, in part because the AI’s confident output discouraged the kind of skeptical review that experienced programmers apply instinctively. If vibe-coded apps with security flaws make it through Apple’s review process and onto users’ devices, the consequences extend well beyond App Store clutter.
Apple’s response in the coming months will be closely watched. The company has historically been willing to make aggressive moves to protect the quality and reputation of its platform — from the original decision to create a curated App Store in 2008 to the ongoing battles over sideloading and alternative app marketplaces in the EU. Tightening review standards, investing in better automated detection of AI-generated submissions, or even adjusting App Store search to more aggressively demote low-quality apps are all options on the table.
So where does this leave the industry? In a strange place. The tools for building software have never been more powerful or more accessible. The demand for mobile applications continues to grow. But the infrastructure designed to ensure quality — human review teams, guideline enforcement, curation algorithms — is buckling under the weight of what those tools have unleashed. Vibe coding has made app development easier. It has not made app development better. And for Apple, that distinction is now a billion-dollar operational challenge with no easy fix.
The flood isn’t receding. If anything, as AI models improve and coding assistants become even more capable, the volume of submissions will continue to climb. Apple’s ability to manage that volume — without sacrificing the standards that have made the App Store a trusted distribution channel — may be one of the most consequential platform governance challenges of the next decade.
WebProNews is an iEntry Publication