OpenAI opened its app store to independent developers on December 17, 2025, and the number attached to it is the reason your feed filled up with “build a ChatGPT app and get rich” videos: 800 million people use ChatGPT every week, and any app that clears review can surface in front of them. That is a bigger built-in audience than the iPhone had when Apple opened the App Store in 2008.
Here is the part the thumbnails skip. Six-plus months in, you still cannot sell a digital product directly inside a ChatGPT app. The payment rails that would let a solo builder charge for a subscription or a one-time download are either restricted to a private beta for large marketplace partners or limited to physical goods. So before you spend a weekend learning the Apps SDK, it is worth understanding what OpenAI actually opened, what it did not, and which kind of builder this is genuinely a fit for right now.
What OpenAI actually opened
In October 2025 OpenAI introduced apps inside ChatGPT alongside a new Apps SDK, with a launch roster of names most people recognize: Booking.com, Canva, Coursera, Expedia, Figma, Spotify, and Zillow. Those were invite-only integrations. The December announcement changed the shape of it: any developer can now submit an app for review, track its approval status, and, once approved, land in a searchable in-product directory that ChatGPT’s 800 million weekly users can browse and that ChatGPT can surface mid-conversation when it is contextually relevant.
The technical foundation matters more than it sounds. The Apps SDK extends the Model Context Protocol (MCP), the same open standard that has been quietly turning into the plumbing of the AI tool economy. If you have already looked at building and selling MCP servers, you are not starting from zero. A ChatGPT app is, in large part, an MCP server with a user interface layer bolted on and a distribution channel wrapped around it. The skill transfers directly, which is a real edge for anyone who spent the first half of 2026 learning that stack.
The directory itself is the headline. For years, the hard part of shipping any AI tool was not building it; it was getting anyone to find it. OpenAI is offering to solve distribution for you, in exchange for building on its terms and passing its review. That trade is the whole story, and whether it is a good trade depends entirely on how you plan to make money.
The monetization gap nobody puts in the thumbnail
Read OpenAI’s own monetization documentation and the current state is unambiguous. There are three ways to take payment through an app, and two of the three are not open to you if you are a solo builder.
The first is external checkout, which OpenAI recommends. Your app links out to your own website, and the customer completes the purchase there. You keep full control of billing, taxes, and fulfillment. You also eat the conversion loss that comes with every extra click and every context switch out of the chat. Anyone who has run a no-code micro-SaaS knows what a hand-off like that does to a funnel.
The second is in-app checkout using a payment method the customer already saved with you. Useful, but it assumes they are already your customer, which most of ChatGPT’s 800 million users are not.
The third, and the only one that feels like a true in-chat storefront, is the ChatGPT payment sheet built on the Agentic Commerce Protocol. That one is in private beta and limited to select marketplace partners. It is not something a first-time submitter gets access to. On top of that, approval right now is scoped to physical goods. The digital products a solopreneur would naturally sell inside a chat, a template pack, a niche tool subscription, a course, are not yet supported through the built-in rails at all.
Put plainly: OpenAI opened the doors to the audience before it opened the cash register to independents. That is not a knock on the strategy. It is just the order they chose, and it changes what the opportunity is.
This is a distribution channel, not a storefront
Once you accept that you cannot reliably charge inside the app yet, the right mental model snaps into focus. A ChatGPT app in mid-2026 is a top-of-funnel channel, not a point of sale. Its job is to put your product or service in front of a colossal audience and route the ones who convert back to a place where you can actually bill them.
That reframes who wins. The builder who already has something to sell, a SaaS product, a service business, a paid community, gains a genuinely new acquisition surface. They can build an app that solves a narrow problem for free inside ChatGPT, prove value, and pull qualified users to their own checkout. This is the same logic behind running a lean AI-first solo business: use the free tool as the lead magnet and monetize the relationship, not the first interaction.
The builder who has nothing to sell yet, who was hoping the app itself would be the product and the money would appear inside the chat, is early. Not wrong, just early. The rails they need are on OpenAI’s roadmap for 2026, including broader Agentic Commerce Protocol support with revenue sharing, but “on the roadmap” and “in your bank account” are different tenses.
There is a related opening that gets less attention: building ChatGPT apps for clients. Local businesses and small brands are going to want presence inside ChatGPT the same way they once wanted a website, then a mobile app, then a chatbot. That is a services play with immediate revenue, and it looks a lot like the white-label AI chatbot and custom AI agent work that already pays. You are not waiting on OpenAI’s payment infrastructure because your client pays you directly.
The comparison worth sitting with
I spent two decades in IT operations before running fractional COO work, and one pattern shows up in every platform gold rush: the platform opens, thousands of builders rush in, and the ones who profit are rarely the ones who moved fastest. They are the ones who understood how discovery and payment actually worked before they invested.
Amazon opened Alexa to third-party skill developers in 2015 with a similar pitch: enormous installed base, open submissions, be early. Tens of thousands of skills got built. Almost none made money, because Alexa had no real discovery mechanism and no clean way to charge inside a voice interaction. The audience was real. The monetization was not, and it never fully arrived before attention moved on.
ChatGPT is not Alexa. The interface is richer, the use cases are broader, and the directory plus contextual surfacing is a far better discovery story than Alexa ever had. But the lesson holds: a giant audience with immature payment rails rewards patience and positioning over speed. The people who win the ChatGPT app store will most likely be the ones who treat the current phase as a land-grab for distribution and relationships, then monetize hard when the rails mature, not the ones burning weekends trying to force a payday the platform is not built to deliver yet.
Who should build now, and who should wait
Build now if you already sell something and want a new front door to 800 million people, if you have MCP or AI workflow experience and the learning curve is short for you, or if you can sell app-building as a service to clients who pay you directly. In all three cases you are not dependent on in-chat monetization, and getting in early buys you directory position and review-process familiarity that latecomers will not have.
Wait, or at least keep it a side experiment, if your entire plan hinges on charging users inside the chat for a digital product. Learn the Apps SDK now so you are ready, ship something free to understand the review process, but do not quit anything on the assumption that in-app digital-goods monetization is live. It is not, for independents, as of this writing.
The honest read on the ChatGPT app store is that OpenAI handed independent builders the single hardest thing to get, distribution to 800 million people, and held back the single most important thing to keep, a way to get paid inside the experience. That gap will close. The builders who position for it now, without pretending it has already closed, are the ones who will be standing in the right place when it does.
Sources: OpenAI, “Developers can now submit apps to ChatGPT” and “Introducing apps in ChatGPT and the new Apps SDK”; OpenAI Developers, “Monetization, Apps SDK”; TechCrunch, “ChatGPT launches an app store”; VentureBeat, “OpenAI now accepting ChatGPT app submissions from third-party devs, launches App Directory”.