Skip to content

OpenAI retires custom GPTs on 11 December. How do you move your team's GPTs?

By Bilal Haider, Founder, Theepa Studio · · 6 min read

Short answer

OpenAI is retiring custom GPTs in ChatGPT and replacing them with plugins. The standard retirement date is 11 December 2026, and Enterprise workspaces with an approved deferral can keep them until 11 February 2027. You migrate each GPT with "Migrate to plugin" in My GPTs, but custom actions, the chosen model, sharing settings and past chats do not carry over, so test and rebuild before the date.

What is happening to custom GPTs, and when?

OpenAI is replacing custom GPTs, the tailored assistants people build in ChatGPT, with plugins. Virtualization Review reported on 28 September 2026 that OpenAI schedules the retirement of custom GPTs for 11 December 2026, and that timing and migration availability can vary by account and workspace. That is 24 days from today.

These are the dates as reported, read on 17 November 2026:

WhoDateWhat happens
Most ChatGPT accounts and workspaces11 December 2026Custom GPTs are retired
Enterprise workspaces with an approved deferral, and FedRAMP workspaces11 February 2027Custom GPTs keep working until then
Affected Enterprise workspaces26 October 2026 (target, now passed)Creating new custom GPTs ends

OpenAI's own migration guide is written for Enterprise admins and GPT creators, and it says timelines for other plans may differ. It also says what retirement means: a custom GPT stops running and leaves the GPT directory. So check the notice in your own My GPTs page, and plan for 11 December unless you have a confirmed deferral.

What is a ChatGPT plugin?

A plugin is a package that can hold skills and MCP servers. OpenAI's developer docs describe a skill as a folder with a SKILL.md file plus optional scripts, references, templates or assets. An MCP server is a connection to another tool, such as your CRM or project board. ChatGPT and Codex share one plugin directory, and plugins work in ChatGPT on the web, desktop and mobile.

If this sounds familiar, it is. Google moved Gemini Gems to skills this month, using the same SKILL.md idea. We covered that in how to save your team's Gems.

What moves to the plugin, and what doesn't?

This is the part that catches teams out. Based on OpenAI's migration guide and Virtualization Review's report:

Part of your GPTWhat happensWhat to do
InstructionsBecome a skillRead the new skill and tidy it
Knowledge filesCome over with the pluginCheck every file arrived
Connected appsCome overReconnect and test each one
Custom actionsDo not transferRebuild with a supported connector or a custom MCP server
Chosen modelDoes not carry overRetest your outputs
Sharing settingsThe plugin starts privateShare it again with the right people
Past conversationsStay with the old GPTCopy out anything you still need
Unpublished draftsOnly the published GPT migratesPublish drafts you want to keep first

OpenAI also warns that "a migrated plugin may respond differently", and recommends retesting prompts, file use, output formats and integrations.

How do I migrate a custom GPT to a plugin?

OpenAI's guide sets out four steps:

  1. 1Convert. Open My GPTs, find the GPT under "Created by me" and select Migrate to plugin. A workspace admin can help.
  2. 2Test. Try the tasks the GPT normally does, plus one harder case. Check skills, files, tools and integrations.
  3. 3Share and verify. Share the plugin with the people or groups who need it. They install it before they can use it, unless an admin installed it for them.
  4. 4Help the team switch. Share the new link, explain what changed, update onboarding and name an owner.

After migration the original GPT becomes read-only. It keeps working until the retirement date, and future edits happen in the plugin.

Worked example: a five-person marketing team moves three GPTs

Here is a typical scenario. A startup marketing team in Dubai has three custom GPTs:

  • Brand Voice: instructions plus a brand guidelines PDF.
  • Proposal Writer: a custom action that pulls deal details from the team's CRM.
  • Social Captions: shared with an outside freelancer.

This is how we would plan their week:

  1. 1Monday: take stock. List each GPT with its owner, files, actions and users. Copy every set of instructions into a shared document as a backup.
  2. 2Tuesday: Brand Voice. Publish any draft changes, then migrate. Run five real prompts from last month through the old GPT and the new plugin side by side. If the tone drifts, add three short examples of good and bad copy to the skill.
  3. 3Wednesday: Proposal Writer. The custom action will not move. Check whether the CRM has a supported connector in ChatGPT. If it doesn't, the action has to be rebuilt as a custom MCP server, which usually needs a developer. Start this one early, because it takes longest. If you want outside help, this is the kind of work our AI workflows service is for.
  4. 4Thursday: Social Captions. Share the plugin with the freelancer and send the new link. They will not get access just because they used the old GPT.
  5. 5Friday: tidy up. Replace old GPT links in onboarding notes and bookmarks, and name one owner per plugin.

One thing the Brand Voice step often shows: the GPT is the only place the brand's tone is written down. If that is true for you, fix the source, not just the assistant. Our brand guidelines checklist covers what a guide should hold, and our brand identity service builds one.

Should we migrate every GPT, or let some retire?

Not every GPT is worth moving. Our advice:

SituationOur advice
Used every week, instructions and files onlyMigrate now, test, share again
Uses a custom actionMigrate first, then rebuild the action. Start here
Public GPT that customers useMigrate, share or publish the plugin, and update every link on your website and emails
Not used in three monthsCopy the instructions into a document and let it retire

If you were also using OpenAI's Agent Builder for bigger automations, that closes on 30 November. See what to switch to.

What should I do before 11 December 2026?

  1. 1List every GPT in My GPTs, with its owner and users.
  2. 2Copy the instructions and download the knowledge files as a backup.
  3. 3Publish any drafts you want to keep.
  4. 4Migrate, starting with GPTs that use custom actions.
  5. 5Test with real prompts and compare against the old GPT.
  6. 6Share the plugins again and update every link.
  7. 7Name one owner for each plugin.

Sources

Bilal Haider

Founder, Theepa Studio

Bilal founded Theepa Studio in 2016. The studio has delivered more than 10,000 brand identities for companies in the US, UK and the GCC.

Questions people also ask

When do custom GPTs stop working?

The standard retirement date is 11 December 2026. Enterprise workspaces with an approved deferral, and FedRAMP workspaces, have until 11 February 2027. OpenAI says a retired GPT stops running and leaves the GPT directory.

Will my custom GPT turn into a plugin on its own?

Don't count on it. OpenAI's guide describes a manual step: open My GPTs, find the GPT under "Created by me" and select "Migrate to plugin". Only the published version migrates, so publish any drafts first.

Do custom actions work in ChatGPT plugins?

Not automatically. Custom actions do not transfer when you migrate. OpenAI says to rebuild them with a supported connector or a custom MCP server, so start with any GPT that uses actions.

Can people who used my public GPT use the new plugin?

Not automatically. A migrated plugin starts private and earlier users do not get access. Share or publish the plugin, then update every link on your website, emails and onboarding notes.

Keep reading

Tell us what you're building.

A 30-minute call. We look at your brand and site together, and you leave with a plan and a fixed price, whether or not you hire us.