Paul
According to SimilarWeb, more than 75 percent of new apps launched in the past year have failed to reach even 1,000 cumulative downloads. Not a thousand active users, not a thousand paying customers, a thousand downloads total, which is a very low bar that failing to clear it means almost nobody found the app. Three quarters of new apps are landing in near silence, and the timing of why is not a coincidence.
I've written before about how building got so easy while getting found stayed hard, and this is that argument with a number attached to it. The thing I want to walk through here is what the data actually shows, why it's happening now specifically, and what it means for anyone deciding to build, because the lesson isn't "don't build," it's something more useful and more specific than that.
What the data shows
The SimilarWeb numbers are stark enough to quote directly. Of Android apps released since February 2025, 75.2 percent have not accumulated even 1,000 cumulative downloads, and only 2.7 percent crossed the 100,000-download threshold. The entire middle of the distribution, everything between 1,000 and 100,000 downloads, accounts for just over 22 percent of releases. So the shape of it is rough: a tiny sliver of apps get real traction, a thin middle gets some, and the overwhelming majority get almost nothing.
Now pair that with the other half of the data, which is the supply side. App launches across iOS and Android are up roughly 50 percent year over year, with growth peaking near 55 percent in early 2026, and the inflection point matches up with almost exactly to when agentic coding tools, the ones that can build and ship a working app with minimal human involvement, became widely available around early 2025. Before that, app release growth was actually in decline, down around negative 35 percent year over year in early 2024. So the market went from shrinking to exploding in the span of a year, and it was flipped because AI was making apps easy and cheap to build.
Put the two halves together and the story tells itself. The number of apps being shipped surged, and the number of users available to download them did not, because there is no equivalent AI tool that manufactures human attention. More supply, flat demand, and the predictable result is that the average new app gets a smaller and smaller slice of a pie that didn't grow.
Why this is happening now
I know there's a lot of AI slop apps out there but it's not that the new apps are necessarily worse. Some of them are genuinely good. The problem is structural, and it comes down to a mismatch between two things that used to be coupled and got decoupled by AI.
Building an app and getting people to use it were always two separate jobs, but for a long time the difficulty of building acted as a natural filter. Building was hard enough that the number of apps entering the market was limited, which meant any given app faced a much more manageable amount of competition for attention. The difficulty of building was, in a weird way, doing distribution a big favor, by gatekeeping the field small enough that getting noticed was at least possible. AI removed that filter. Now anyone can build, building is fast, easy, and cheap, and the field is enormous, but the audience on the other side is exactly as finite as it ever was. The filter that used to limit supply is gone, and nothing replaced it on the demand side.
SimilarWeb says that this isn't a new dynamic, since most app releases have always struggled for visibility, but what's new is the sheer volume of low-traction apps entering the market simultaneously. That's the key insight. App store obscurity isn't new. What's new is how many people are now experiencing it at once, because the building constraint that used to keep them from ever shipping is gone, so they ship, and then they discover that shipping was never the hard part.
What nobody wants to hear
Here's the uncomfortable takeaway. The ability to build something was never the valuable part. It felt valuable when it was scarce, when being able to make an app put you in a small club that could do something most people couldn't. AI dissolved that scarcity, and in doing so it revealed that the building was never where the value actually lived. The value was always in the part that's still hard, which is getting people to know your app exists and care about it.
This is why "I built an app" means nothing now as a business statement, the same way "I wrote some words" means nothing as a publishing statement. Of course you did. So did everybody else, and the AI helped all of you do it just as fast. The 75 percent failing to crack a thousand downloads aren't failing because they built badly, they're failing because they built without a plan for the part that AI doesn't touch, and that part is now the entire game. When supply explodes and demand stays flat, distribution stops being one factor among many and becomes the only factor that separates the 2.7 percent from everyone else.
I find this genuinely clarifying rather than depressing, because it tells you exactly where to point your efforts. If building is no longer the bottleneck, then spending more of your energy on building is spending it on the wrong thing. The hours AI frees up by making the build fast are not hours to build more apps. They're hours to do the distribution work that the 75 percent skipped, the audience-building and the positioning and the actual reaching of humans, which is harder and slower and far less fun than building but is now what determines whether anything you make gets seen.
So what do you actually do?
The move is to flip the order in which you think about a project. The instinct, especially now that building is so easy, is to build it first and figure out distribution later, and the SimilarWeb data is essentially a graph of what happens when a few hundred thousand people all follow that same instinct. The better order is to figure out how your app will get found before you build it, to treat distribution as the first question rather than the last, because if you don't have a believable answer to "how will anyone find this," then building it faster just gets you to the silence sooner.
That doesn't mean don't build. It means build with your eyes open about what the build does and doesn't get you. The build gets you a product. It does not get you users, it never did, and now that everyone can build, the gap between having a product and having users is wider and more crowded than it has ever been. The founders who end up in the 2.7 percent are not the ones who built fastest. They're the ones who treated the finding as the real work and the building as the easy part, which, thanks to AI, it now genuinely is.
So if you're about to start something, ask the hard question first. Not "can I build this," because the answer is yes and the answer being yes is worth nothing. Ask "when this exists, what is the actual way someone discovers it and decides to care," and if you don't have an answer, that's the thing to go work on before you get Claude Code to write a line of code. The 75 percent are a warning, not about building, but about building as if building were enough. It stopped being enough the moment it stopped being hard.
Sources
OfficeChai: Agentic Coding Has Led to a 50% Increase in Apps but Most Are Finding Very Few Users - SimilarWeb's data that 75.2% of Android apps released since February 2025 have under 1,000 cumulative downloads, only 2.7% crossed 100,000, app launches up roughly 50% year over year tracking to the rise of agentic coding tools, and the prior decline in app releases before early 2025.
Similarweb: The App Explosion After AI and What It Means - The original SimilarWeb analysis of how AI-accelerated app creation is reinforcing winner-takes-most dynamics across the mobile economy.
