# What Did OpenAI Actually Ship, and Why Call It an App Store Alternative?

TechCrunch's Sarah Perez reported on September 29, 2026 that OpenAI is "building out the pieces of an alternative to the traditional app store model," with ChatGPT becoming a surface where software is discovered and used by both humans and AI agents. The critical word in that sentence is "pieces" — this is an assembly story, not a launch story. The distinction matters. Apple did not beat Nokia with a single product; it beat Nokia with a store, a payment rail, and a developer agreement that made the store the default. OpenAI is doing the same thing in reverse order: it already has the users and the payment relationship (ChatGPT subscriptions), and it is now bolting on the discovery and invocation layer. My read: the agent-readable listing is the actual innovation here. A traditional app store assumes a human browses, taps, and installs. An agent surface assumes software is described in a way a model can select and call. That is a different indexing problem, and neither Apple's App Store nor Google Play was architected for it.

# Why Does Agent-Readable Software Change the Discovery Economics?

According to TechCrunch, the new surface serves "people and AI agents alike." That phrasing is doing enormous work. If an agent can find and invoke a tool without a human ever opening a store page, then the store page — the thing Apple charges for placement on — loses its function as a toll booth.

Apple's App Store economics depend on the storefront being the only practical path between developer and user. Every commission, every ranking algorithm, every editorial feature exists because that path is narrow. An agent that resolves a user's intent to a tool directly collapses the width of that path to near zero. This is inference, not reported fact — TechCrunch does not quantify the revenue impact, and OpenAI has not published commission terms. But the structural logic is hard to escape: distribution layers capture rent only while they remain the shortest path.

# How Do OpenAI, Apple, and Google Compare on the Things That Actually Matter?

Dimension	OpenAI / ChatGPT	Apple App Store	Google Play
Primary discovery mode	Conversational + agent invocation (reported Sept 2026)	Human browse, search, editorial	Human browse, search, algorithmic
Developer commission	Not publicly disclosed	15–30% standard tier	15–30% standard tier
Agent-readability of listings	Core design goal per TechCrunch	Not architected for it	Not architected for it
Payment rail lock-in	OpenAI subscription relationship	In-app purchase mandate	Google Play Billing
Regulatory exposure	Low so far	High (EU DMA, US litigation)	High (EU DMA)
Verdict	Wins on cost of discovery; loses on developer lock-in today	Defends, but on borrowed time	Most exposed of the three

# Who Actually Gains From a Third Storefront?

Small developers gain the most and know it least. A solo developer paying Apple 30% on a $10 subscription keeps $7. The same developer listing inside ChatGPT keeps whatever OpenAI's undisclosed terms allow — and if those terms are zero or near-zero, the arbitrage is obvious. Apple loses slowly and then all at once. Services revenue is Apple's growth story, and App Store commissions are its highest-margin component. A distribution alternative does not need to win; it only needs to make the commission negotiable. Google is the most exposed. Google Play's commission base is more Android-fragmented, and Google has already been forced into billing concessions in multiple jurisdictions. A third surface with no commission is a cleaner pitch to the same developers Google is already losing arguments with.

# What Is Still Unproven Here?

Almost everything operational. TechCrunch's Perez framed this as OpenAI "building out the pieces" — which is a statement about direction, not completion. There is no reported commission schedule, no developer agreement terms, no disclosed revenue share, and no confirmed timeline for general availability. The honest uncertainty: OpenAI has not demonstrated it can police a software marketplace. Apple's commission buys curation, fraud review, and payment dispute handling. If OpenAI charges nothing, it must fund those functions from subscriptions — a subsidy that holds only while ChatGPT's growth holds.

# Predictions

1. OpenAI will publish developer terms with a revenue share below 15% before Q3 2027. The entire strategic value of the surface depends on undercutting Apple and Google, and a zero-commission model is unfundable at scale — so expect a number, not a giveaway.
2. Google will announce a reduced Play commission tier for agent-invoked software within 12 months. Google has already conceded on billing in multiple jurisdictions and is structurally more vulnerable than Apple to a distribution alternative.
3. Apple will not change its standard commission in 2027, but will expand its small-business tier eligibility. Apple's pattern is to concede at the edges rather than the center, and the edge is where the developer defection risk is highest.

# Article Summary

• TechCrunch reported on September 29, 2026 that OpenAI is assembling an app-store alternative inside ChatGPT — an assembly story, not a launch.
• The genuinely new element is agent-readable listings, which attack the store page's function as a toll booth rather than its design.
• Small developers gain the most from a zero- or low-commission surface; Google is more exposed than Apple because it has already conceded on billing elsewhere.
• OpenAI has published no commission terms, no developer agreement, and no availability timeline — the economics are entirely unproven.
• The real test is whether a marquee iOS developer lists a ChatGPT-native product that undercuts its App Store pricing — I expect that by end of 2027.