Timing: budget weeks, not days
Two dates matter. The created date is encoded in the app's id; the store's field for it is called launched_at, but it marks when the id was made, not a launch. The go-live date, labelled went live, is the day the app first appears in the directory. The gap between them is the launch lag.
The sample is small, so read it as a range, not a promise. 20 apps in our listing sample went live from Oct 7 to Oct 13, outside the Oct 8–9 batch, and 19 of them carry a created date. The fastest took 3.7 days and the slowest 191.1. The middle half took 13.7 to 56.2 days. 6 went live in under two weeks; 5 took more than 60 days.
Two things this can't show. It counts apps that went live in one week, so it can't see apps still waiting (right-censoring). And it leaves out the 410 listings in our listing sample that appeared together on Oct 8–9 (1,117 store-wide in our Oct 13 analysis), with ids created as early as Sep 12, 2025. That migration batch would swamp a sample of 19; the report on the Oct 8–9 batch covers it.
Category: know the list you're entering
Your category decides which list ranks you and which home shelf you can reach. Shelves are 9-app runs of their list, so the size of the list is the size of the field. Productivity has 1,400 apps and Business & Operations 1,137; at the other end, Communication has 53, Security 46 and Scientific Research 28.
Pick the category that describes what the app does, then read your rank against that list, not the whole store. A smaller list means fewer apps to pass before the shelf. How to get on a ChatGPT home shelf has the shelf-by-shelf numbers.
Countries: check where you're listed
The directory's order is the same in every country, but the set of apps isn't. Finance and Healthcare are the categories least often listed everywhere: 85 of 100 and 86 of 99 of their top-100 apps appear in all 42 countries with data, against 100 of 100 in Productivity. 32 apps from US top-100 lists are listed only in the US, 11 of them in Healthcare and 9 in Finance, including Intuit QuickBooks, Webull and SaveHealth.
We don't see why an app is missing in a country; we see that it is. If your app handles money or health, expect gaps and check each country in the first week. Data is not available for six countries (Belgium, Bulgaria, Ireland, Romania, Serbia and Slovenia), so for those we show unknown, not missing. Why country availability matters and Availability across 48 countries have the country-by-country numbers.
Listing: the first thing people see
On a shelf or a list, people see a name, an icon and one line. On the app's page they also see the longer description, starter prompts and screenshots. Of the 29 ChatGPT apps OpenApps tracks in full (19 of them in Popular), 25 show starter prompts and 9 show screenshots. Prompts are the norm at the top; screenshots aren't.
Write the one line for someone deciding in two seconds: what the app does, in their words. Make starter prompts tasks a new user can run on the first try. We keep every version of a listing, so every listing change after launch, yours or a competitor's, shows up in the Listing changes feed.
Your MCP server
The store publishes no MCP server URLs for ChatGPT apps, but the servers OpenApps tracks still show what to copy. In our wider MCP dataset, which covers the open servers we track across every store and not only the ones behind ChatGPT and Claude listings, 91% of tools set readOnlyHint, but only 39% declare an output schema. Before launch:
- Set readOnlyHint and destructiveHint truthfully on every tool.
- Declare output schemas for tools that return records.
- Keep the tool count small; the median open server behind a live listing has 6.
- If your app is also a Claude connector, its OpenApps app page shows its MCP status and, unless it asks for sign-in, its tools.
What MCP tool design teaches has the full numbers and a checklist.
The first two weeks
- Watch for your go-live. Your go-live date is the day you first appear in the directory. New launches lists apps by the day they went live, and the Launch lag page tracks the gap.
- Read cohort rank, not raw rank. In the 24 hours to October 13, 2026, 224 rows changed position in ChatGPT's ranked top-100 lists, and by cohort rank no app moved. A new app's raw rank will drift as others arrive; cohort rank shows whether you moved.
- Check the shelf, not only the rank. Your category shelf is #1 to #9 of your list once Popular's top 6 are taken out.
- Check every country. Look at all 42 countries with data, not only your own.
- Give the ranks time. A two-week-old app has two weeks of ranks. Judge its trend over a month, not its first days.
Create the app well ahead: 5 of the 19 apps in our listing sample took more than 60 days from creation to going live. Announce when you see the app in the directory, not before.