Guides & TutorialsAI News

Custom GPTs Retire Dec 11: Move to a Project Now

OpenAI plans to retire custom GPTs on December 11, 2026. Here is which dates are reported, what a migration keeps and drops, and how to pick a Project or a plugin.

Toolbit AI - Team
4 min read
Custom GPTs Retire Dec 11: Move to a Project Now

Custom GPTs are scheduled to retire on December 11, 2026, according to SiteSpeak and other outlets citing OpenAI's retirement FAQ. Copy your GPT's instructions and files out now, then rebuild it as a ChatGPT Project or migrate it to a plugin.

What happened: OpenAI is retiring custom GPTs on Dec 11

OpenAI announced the retirement on September 11, 2026, with a migration path to plugins, according to SiteSpeak's timeline. We couldn't open OpenAI's own help pages, so every date here is secondhand.

  • What changed: custom GPTs are scheduled to stop running on December 11, 2026.
  • Who's affected: anyone who built or relies on a custom GPT; Enterprise workspaces have the most detailed dates.
  • What to do: save the instructions and files, then pick a Project or a plugin per GPT.

Timelines vary by account, so check the notice in your own account or workspace first.

Unconfirmed: that personal plans can no longer create GPTs, and that GPT chats can't move into a Project (plan as if they stay behind).

Which dates are reported, and which are only planned

The December 11 date appears in every outlet we checked, while the Enterprise dates come from one timeline. SiteSpeak lists these.

DateWhat happensStatus
Sep 11, 2026Retirement announcedHappened
Sep 22, 2026Migrate-to-plugin rolls outRolling out
Oct 26, 2026Enterprise creation of new GPTs endsWas scheduled
Dec 11, 2026Custom GPTs retireScheduled
Feb 11, 2027Retirement for Enterprise workspaces with an approved deferralConditional
Custom GPT retirement timeline: announced Sep 11, plugin migration from Sep 22, Enterprise creation ends Oct 26, retirement Dec 11, 2026, deferred Enterprise date Feb 11, 2027

Your existing GPTs keep running until the retirement date that applies to you.

What a plugin migration keeps, and what it drops

A plugin migration moves the instructions, connected apps and knowledge files, but not custom actions, sharing settings or chats. The Paperless Agent and SVA describe the split this way:

In your custom GPTAfter migration to a pluginWhat to do
InstructionsMove overRetest; a migrated plugin may respond differently
Knowledge filesMove overReview the copies
Connected appsMove overCheck each app still connects
Custom actionsDo not moveRebuild with a supported connector or a custom MCP server
Sharing settingsDo not carry overShare the plugin again
Past conversationsStay behindCopy what you need into a note; the old GPT stays usable until retirement
Conversation starters and selected modelDo not carry overSave the starters separately

SiteSpeak adds two catches: "Migration does not install the plugin, not even for you," and the original GPT becomes read-only and can't be deleted. So finish any edits in the GPT before you migrate.

Copy these five things out of your GPT today

Copying the parts by hand is the only backup you control, whichever route you pick. First, list every GPT you rely on, who built it and in which account or workspace. Flag any with custom actions; they take longest to rebuild. Then open the GPT's editor and save:

  1. The instructions, verbatim, in a document you own. Writing them as reusable context engineering means they outlive the next container too.
  2. Every knowledge file you uploaded.
  3. The conversation starters, since they don't carry over.
  4. A few test prompts and answers, including one hard case, to compare the replacement against.
  5. Connected apps and custom actions: which API each action calls, its schema and how it signs in.

Rebuild it as a Project in six steps

A Project holds chats, files and instructions together, so it is the closest rebuild if your GPT is mostly a prompt plus files. Our ChatGPT Projects guide has the current menu paths.

  1. Create a Project named after your GPT.
  2. Paste the saved instructions into the Project's instructions, then cut lines about GPT-only features.
  3. Upload the knowledge files. Otio, citing OpenAI's file-upload page, reports 5 files per Project on Free, 25 on Go and Plus, and 40 on Edu, Pro, Business and Enterprise.
  4. Pick the memory setting, covered in the guide, before you add anything confidential.
  5. Start fresh on chats: past GPT conversations stay behind, so upload a note of key conclusions instead.
  6. Run your saved test prompts and adjust the instructions until the answers match.

Claude and Gemini have similar features; see Projects compared or our wider Claude vs ChatGPT guide.

Pick a Project or a plugin with one rule

Choose by what the GPT actually does:

  • Mostly instructions and files: rebuild it as a Project.
  • Needs apps or skills in a Business or Enterprise workspace: migrate to a plugin. Steps reported by The Paperless Agent: open My GPTs, select the GPT, choose Migrate to Plugin, read the warning, confirm. Then open it from a new chat with an @ mention or the + menu. Features that use custom actions won't work, but you can migrate without them.
  • Has custom actions: plan a rebuild with a connector or MCP server now. Start with what MCP is and what a skill is.
  • Someone else's GPT: having permission to use it doesn't let you migrate it, Florida Realtors reports. Ask the creator whether a replacement is coming, and for anything you depend on, build your own Project version from the prompts you use.

Migration and plugin access may differ by account or workspace. One forum user quoting the FAQ says personal accounts should follow an "in-product notice."

Share this article

Related articles

Continue exploring similar guides and insights