Getting your app into ChatGPT: the channel nobody is fighting for yet
OpenAI opened third-party app submissions in 2026. Roughly 800 million weekly users, minimal competition, and a directory that has already been reorganised once.
the short answer: openai began accepting third-party app submissions in early 2026, which puts a distribution surface in front of roughly 800 million weekly ChatGPT users, larger than the apple app store’s installed base. competition is currently minimal. it is also the least stable channel on this site: the directory was already reorganised in july 2026, so verify everything before you build against it.
Why this is worth attention
every canonical growth story people still quote came from the same condition: a large new surface with few participants and no established playbook. airbnb on craigslist, dropbox in an era of scarce storage, the early app store.
that condition is rare and short-lived, and it is what exists here right now. the difference from a normal app store launch is that discovery inside an assistant is conversational, so the unit of competition is whether the model reaches for your app when somebody describes a problem, not whether your icon is visible in a list.
What it actually is
apps built on the model context protocol, submitted to openai and surfaced inside ChatGPT. the directory arrangement changed in july 2026, with the app directory folded into the plugin directory and apps continuing as the integrations underneath.
that reorganisation matters practically: most published guidance about this channel predates it and is now partly wrong. anything you read about ChatGPT app distribution, including this post, needs its date checked.
Who this is and is not for
good fit: products that do a discrete, describable job. a user says “help me do X”, and your app is the thing that does X. tools, converters, lookups, generators, anything where the request maps cleanly onto a capability.
bad fit: products whose value is the interface, the community, or sustained engagement. if the point of your app is that people open it every day, being invoked inside somebody else’s chat does not serve that.
worth thinking about carefully: anything where you need the customer relationship. distribution inside an assistant means the assistant owns the user, and you are a capability it calls. that is the same trade as any platform, taken further.
The realistic assessment
volume today is small. this is a land grab bet, not a traffic source. anyone telling you it is a meaningful channel in 2026 is guessing.
the maintenance burden is the highest of anything i would recommend. the surface changed once already this year. building here means committing to re-verifying quarterly.
the asymmetry is what makes it interesting. the cost of being early is a few days of engineering. the cost of being late, if it becomes a real channel, is competing for position in a directory that already has incumbents.
How I would sequence it
do not do this instead of the things that work. do it after.
if your creative throughput is below the benchmark of one new ad per $3,000 of monthly spend, fix that first, because it has a known return. if your first-renewal rate is broken, fix that first, because nothing else matters until it is. this is a bet you place with spare capacity, not with the capacity you need elsewhere.
then: build the integration, submit it, and treat any resulting traffic as an option rather than a plan.
The broader point
the surfaces where distribution is cheap keep moving, and they are cheap precisely because nobody has figured out how to work them yet. search was that once. the app store was that once. social was that once.
the operators who benefit are the ones already running a testing system, because a new surface is just another test with an unusually good prior. the ones who do not benefit are the ones who read about it, build for it exclusively, and discover the surface changed underneath them.
if you want help deciding whether this is a sensible bet for your product specifically, let’s talk.
sources: openai announcements on third-party app submissions, 2026 · reporting on the july 2026 app directory reorganisation · published ChatGPT weekly user figures. written august 2026; re-verify before building, this surface has already changed once this year.