OpenAI's GPT Store opens with 3 million custom GPTs and a revenue plan that isn't finished

From 3 million private builds to a ranked catalog

OpenAI says users created over 3 million custom versions of ChatGPT in the two months since GPTs launched. The GPT Store is the answer to the obvious next question: how does anyone find the useful ones among three million.

The store rolls out first to ChatGPT Plus, Team, and Enterprise users at chatgpt.com/gpts. Discovery is organized two ways — a community leaderboard of popular and trending GPTs, and weekly editorial picks. The categories named are DALL·E, writing, research, programming, education, and lifestyle.

The first featured GPTs come from recognizable partners: AllTrails for trail recommendations, Consensus searching 200M academic papers, Khan Academy's Code Tutor, Canva for design, a Books recommendation GPT, and CK-12's Flexi tutor. The mix signals what OpenAI wants the store to look like — branded utilities from known names, not just hobbyist experiments.

A publishing gate with human and automated review

To list a GPT, a builder has to save it for Everyone, verify a Builder Profile with a name or verified website, and comply with usage policies and brand guidelines. GPTs shared only via link stay out of the store.

The identity verification requirement matters. By tying listings to a verified name or domain, OpenAI attaches accountability to every public GPT before it appears. That is a deliberate friction on a platform where the barrier to building is otherwise near zero — no coding required.

OpenAI describes a new review system layered on top of existing product safety measures, combining human and automated review, plus user reporting. For a catalog scaling into the millions, the automated tier is doing the heavy lifting; the human review is presumably reserved for edge cases and appeals.

A revenue program announced before its rules exist

OpenAI says a GPT builder revenue program will launch in Q1, with US builders paid first, based on user engagement. The company adds that it will "provide details on the criteria for payments as we get closer."

In Q1 we will launch a GPT builder revenue program. As a first step, US builders will be paid based on user engagement with their GPTs.Montana Labs

This is an incentive announced without its mechanics. Engagement-based payouts define the game builders will optimize for, and OpenAI is asking people to build toward a scoring system it hasn't published. The gap between announcing the program and specifying the criteria is where builder trust — and the risk of engagement-gaming — will be decided.

The workspace controls are the enterprise story

Alongside the public store, OpenAI gives Team customers a private section holding GPTs published only to their workspace. Enterprise gets admin controls that go further: choosing how internal-only GPTs are shared and which external GPTs are allowed inside the business.

OpenAI also states that conversations with GPTs on Team and Enterprise are not used to improve its models — the same data posture it applies across those plans.

The implication for applied teams: the GPT Store is two products wearing one name. The public leaderboard is a consumer marketplace competing on engagement, while the private workspace with allowlist controls is the version enterprises will actually adopt. The consumer catalog gets the headline; the admin gating over which external GPTs can run inside a company is what determines whether GPTs enter regulated workflows at all.