Tony Wang9 min readThe AI App Graveyard: 8,505 Apps, and the Median Has 10 Ratings
We pulled 8,505 AI apps off the App Store — 5.7× our first study. The median has 10 ratings, 87% have under 1,000, and the flood is still accelerating.
A few weeks ago we pulled 1,490 App Store apps that call themselves "AI" and found a brutally top-heavy market: a few giants, a long tail of clones. The most common pushback was fair — your sample is too small.
So we went back and pulled 5.7× more: 8,505 unique AI apps, gathered by paginating ~100 AI search terms across every sub-category and brand until the App Store stopped surfacing new ones. It's the near-complete searchable universe of AI apps. The findings didn't soften with scale. They got worse — because a bigger net mostly catches more of the dead.
(Method note, as before: the stores rate-limit and fingerprint plain scrapers, so we read them through a structured web-data API. US storefront, June 2026. "AI app" is a strict tag — explicit AI/GPT/assistant signals or a named AI brand — over the search results; caveats at the end.)
Picture the App Store's AI apps as 100 squares
Eighty-eight of them are ghosts. Each square below is about 85 apps, colored by how many ratings it has ever collected:
Each square ≈ 85 apps. The grey field — 88 of every 100 AI apps — has fewer than 1,000 ratings; only 8 apps in the whole set have more than a million.
That grey field is the whole story. Most "AI apps" aren't competing for users — they were never found by any.
A median of 10 ratings
Here's what scaling the sample did to the typical app. When we had 1,490 apps, the median one had 171 ratings. At 8,505, the median is 10:
median ratings
median ratings
The median dropped 17× not because 2026's apps are worse than the ones we saw first, but because the first, smaller pull skimmed the visible head. Dig deeper and you find what's actually there: a near-bottomless tail of apps with single-digit ratings.
From 11,000 results to 8 giants
Run the whole thing as a funnel. About 11,000 apps surface for AI searches; 8,505 are genuinely AI; of those, a few hundred have a real audience and eight have a massive one:
Show the flows
| AI search results → Genuinely AI | 8,505 (43.7%) |
| Genuinely AI → Under 1,000 | 7,480 (38.4%) |
| AI search results → Not AI | 2,453 (12.6%) |
| Genuinely AI → 1k – 10k | 705 (3.6%) |
| Genuinely AI → 10k – 100k | 252 (1.3%) |
| Genuinely AI → 100k – 1M | 60 (0.3%) |
| Genuinely AI → Over 1 million | 8 (0%) |
The flood is accelerating, not slowing
You might assume the AI-app gold rush peaked with ChatGPT's debut. It didn't. By launch year, 2025 produced more AI apps than 2023 and 2024 — and 2026 is on pace to break the record:
The app farms
Most of these apps come from someone's only app — ~6,600 developers for 8,505 apps, 83% of them one-and-done. But a handful of studios have industrialized the gold rush, each shipping a dozen or two near-identical AI apps:
What they all are
Cluster the 8,505 by what each app actually does and it's a narrow menu — chatbots, photo generators, and an enormous pile of AI companions:
How concentrated is it?
Rank all 8,505 apps by rating count and the inequality is near-total — a Gini of 0.98. Five apps hold 38% of every rating; a hundred hold 82%; everything else shares the scraps:
The most-rated "AI" apps are the same incumbents-with-a-feature we saw last time — ChatGPT (8.0M), Canva (3.4M), DuckDuckGo (2.3M), Gemini, Gauth, Microsoft Copilot, Picsart. The 8,400 apps behind them are not, mostly, competing with those. They're invisible.
How we did this (and the caveats)
We gathered the corpus by paginating ~100 AI-intent search terms through a structured App Store API (US storefront, June 2026), deduping to 10,958 unique apps, then strict-tagging 8,505 as genuinely AI (explicit AI/GPT/assistant signals or a named AI brand — not the bare word "chat", which over-counts messaging apps). Caveats, plainly: this is the discoverable search universe, not a true census, so the real tail is even longer and the inequality even higher than the Gini 0.98 we measured — every figure here is a lower bound. Rating count is a proxy for scale, not exact installs. The strict tag still catches incumbents that bolted AI on (DuckDuckGo, Canva, Goodnotes). And we make no claim that the long tail is bad — a median of 10 ratings means undiscovered, which is a very different thing from a bad app.
The deeper read is simple: the first study undercounted the graveyard. At 1,490 apps the median had 171 ratings; at 8,505 it has 10. Keep digging and that number keeps falling — because the App Store's "AI" layer is overwhelmingly apps that nobody ever opened.
This is the kind of thing we read for a living. Crawlora is a web-data API for AI agents and pipelines that returns normalized JSON for the App Store, Google Play, search engines, marketplaces, and social platforms — handling proxies, rendering, and anti-bot — billed pay-on-success. The crawl behind this post was ~100 API calls and 16 minutes. We've since indexed the whole catalog — 4.1M apps across the App Store and Google Play — browsable in our mobile app dataset, and track the daily winners and losers in the App Store + Google Play charts dataset.
Read the App Store and 40 other platforms as clean JSON
Crawlora turns app stores, search engines, marketplaces, and social platforms into normalized JSON — handling proxies, rendering, and anti-bot — and bills pay-on-success. 2,000 free credits a month, no card.
Frequently asked questions
How many AI apps are on the App Store?
We found 8,505 discoverable AI apps (June 2026, US App Store), gathered by paginating about 100 AI search terms — 5.7 times more than our earlier 1,490-app study. The true number is higher still; this is the searchable universe, not an exhaustive census.
What is the typical AI app like?
Forgettable. The median AI app has just 10 ratings, 87% have fewer than 1,000, and only 8 of the 8,505 have more than a million. Most were never found by anyone — ignored, not necessarily bad.
Is the AI app boom slowing down?
No — it is accelerating. By launch year, 1,434 of these apps shipped in 2023, 1,798 in 2024, and 2,890 in 2025; 2026 is already past 1,400 at the half-year mark. 89% launched in 2023 or later.
Who makes all these AI apps?
About 6,600 developers, 83% of them shipping a single app. But a few app farms have industrialized the gold rush — one studio (ZipoApps) ships 26 distinct AI apps, several others 10 or more.
How was this measured?
By paginating about 100 AI-intent search terms through a structured App Store API (US storefront, June 2026), deduping to 10,958 unique apps, then strict-tagging 8,505 as genuinely AI. Rating count is a proxy for scale, and the figures are a lower bound on the true long tail.