Custom GPT Migration: How to Move to ChatGPT Plugins Before Retirement

Custom GPTs are being retired. Learn how to migrate your Custom GPT workflow to a ChatGPT Plugin, see what transfers, identify what must be rebuilt, and test the replacement before switching.

Everyday AI Guides · Updated September 2026

If you built a Custom GPT that you still rely on, custom GPT migration is now something worth planning before the retirement date. OpenAI says Custom GPTs are scheduled to retire on December 11, 2026, although migration availability and some workspace timelines can differ by account.

The good news is that you do not necessarily need to rebuild everything from scratch. OpenAI’s migration flow can move important parts of a published GPT into a ChatGPT Plugin. The important catch is that not everything transfers, and a migrated Plugin may not behave exactly like the GPT it replaces.

This guide walks through how to migrate a Custom GPT to a Plugin, what transfers, what does not, how to handle Custom Actions, what to check when Migrate to plugin is missing, and how to test the replacement before depending on it.

Quick answer

When migration is available for your account, open My GPTs, find a GPT under Created by me, and choose Migrate to plugin. OpenAI says the latest published version is used for migration. Instructions can become a skill, knowledge files can become reference files, and connected apps can carry into the Plugin. Custom Actions, old conversations, the selected model, and existing sharing settings do not transfer automatically.

Best practice: finish important edits first, publish the version you actually want to migrate, save several test prompts, migrate, and then compare the new Plugin against the original GPT before switching your workflow.

When Are Custom GPTs Retiring?

OpenAI currently says Custom GPTs are scheduled to retire on December 11, 2026. Certain affected Enterprise workspaces with an approved deferral may have a later retirement date, so organizations should follow the notice shown inside their own workspace rather than assuming one date applies everywhere.

Existing GPTs remain usable until their applicable retirement date. However, once you migrate a GPT, the original becomes read-only. You can still use it until retirement, but future changes need to be made to the replacement Plugin.

Important before you migrate Migration uses the GPT’s latest published version. Draft changes and unpublished edits do not transfer. Finish any essential edits and publish the version you want to preserve before starting migration. Publishing does not mean you must make the GPT public.

What Transfers During Custom GPT Migration?

A successful custom GPT migration is not a perfect clone. Think of it as moving the reusable parts of your GPT into a different workflow system.

Custom GPT component Migration result What you should do
Instructions Transfers They become reusable instructions inside a Plugin skill. Review them after migration.
Knowledge files Transfers They become reference files. Confirm the files are present and used correctly.
Connected apps Can transfer Check account authorization, permissions, and whether the required actions still work.
Custom Actions Do not transfer automatically Replace them with an available app or rebuild the integration where supported.
Selected model Does not transfer Test output quality instead of assuming the Plugin will behave identically.
Existing GPT conversations Do not transfer Save important information or examples from old conversations separately.
Sharing settings Do not carry over automatically Review access and share the replacement with the intended users after testing.
Draft or unpublished edits Do not transfer Publish essential changes before migration.

OpenAI recommends reviewing the migrated instructions, files, examples, tools, and integrations before relying on the replacement. That is important because even when the content transfers correctly, the Plugin may select tools or follow instructions differently from the original GPT.

Before You Migrate: Use the Designs24hr SAFE Check

The easiest way to lose something during migration is to start before documenting what actually makes your GPT useful. Use this simple pre-migration framework first.

S Save important instructions

Record the current instructions, important examples, and any behavior you cannot afford to lose.

A Audit Actions and apps

Identify every external service the GPT depends on, especially Custom Actions that will need separate work.

F Finish essential edits

Complete and publish the version you actually want migrated. Unpublished edits are not the version migration uses.

E Establish test examples

Save several prompts and expected outputs so you can compare the Plugin with the original GPT after migration.

SAFE is a Designs24hr editorial framework for organizing the migration process. It is not an OpenAI product feature.

What should you save before migration?

  • The GPT’s important instructions and rules.
  • A list of knowledge or reference files you expect it to use.
  • The names of connected apps or services.
  • Any Custom Actions and what each action currently does.
  • Three to five familiar prompts that reliably test the workflow.
  • At least one difficult or edge-case prompt.
  • The expected audience and current sharing setup.
  • Examples of any output format that must remain consistent.

How to Migrate a Custom GPT to a ChatGPT Plugin

The exact interface can vary by account and workspace, but OpenAI’s current migration guidance uses the following basic workflow.

Step 1: Open My GPTs Go to your GPT management area and find the GPT under Created by me. You generally need to be the creator of the GPT or have the appropriate workspace administrator access.
Step 2: Confirm the GPT is ready Finish any important edits and make sure the version you want to preserve has been published. The migration flow uses the latest published version rather than unpublished draft changes.
Step 3: Choose Migrate to plugin When the feature is available for your account, select Migrate to plugin for that GPT and review the migration details shown by ChatGPT.
Step 4: Review what was created Check the migrated instructions, reference files, connected apps, examples, and available tools. Do not assume that seeing the Plugin means the migration is complete.
Step 5: Rebuild anything migration could not move Custom Actions require separate attention. If the workflow depends on an external service, determine whether an available app provides the required action or whether a custom integration needs to be rebuilt.
Step 6: Test before sharing Run familiar prompts and one harder case. Check instructions, reference material, output structure, tools, files, integrations, and access.
Step 7: Share the replacement only after it works A migrated Plugin starts private. Existing GPT users do not automatically gain access to the replacement, so verify sharing and installation separately.
New to the 2026 Plugin system? The word “Plugin” means something different in the current ChatGPT system than it did in the original plugin era. Read our ChatGPT Plugins in 2026 guide for a beginner-friendly explanation of Plugins, skills, apps, and the Plugin Directory.

What If “Migrate to Plugin” Is Missing?

If you cannot see Migrate to plugin, do not immediately assume your GPT is broken. OpenAI says migration access can become available at different times and can depend on the account or workspace you are using.

1. Check the account or workspace where the GPT was created

If you use several ChatGPT accounts or workspaces, make sure you are viewing the one where the GPT was actually created. Permissions and migration availability follow that account or workspace.

2. Confirm that you own the GPT

Being able to use someone else’s GPT does not give you the ability to migrate it. The creator—or an eligible workspace administrator where supported—needs to handle migration.

3. Check whether Plugins are available

Plugin availability can differ by account and workspace. In managed environments, an administrator may also control whether Plugins can be used.

If the whole Plugin Directory is unavailable rather than only the migration button, use our dedicated guide to fix ChatGPT Plugins not showing.

4. Check whether the GPT is published

Migration is based on a published GPT version. Drafts cannot simply be migrated as the current replacement. Publishing the GPT does not require making it publicly discoverable.

5. Check your workspace permissions

In managed workspaces, Plugin use, migration, sharing, or publishing can depend on administrator-controlled permissions. If the migration option should be available but is still missing, ask the workspace administrator to verify that Plugins are enabled and that your account has the required role.

Do not recreate the GPT just because the button is missing Migration access can roll out separately from the retirement announcement. First verify account, ownership, publication status, Plugin access, and workspace permissions.

What Happens to Custom Actions?

This is one of the most important parts of custom GPT migration. OpenAI says Custom Actions do not transfer through the migration flow. If your GPT depends on a Custom Action to reach an API or outside service, that part of the workflow can stop working in the replacement until you build an alternative.

Start by identifying what the Action actually does. Then check whether an available app can provide the same supported read or write capability. More technical workflows may require a separately built integration, such as a supported custom MCP server, depending on the account and workspace.

Do not assume a rebuilt integration is equivalent Even if you successfully reconnect the same service, the new integration may expose different actions, permissions, data, confirmation requirements, or limits. Test the exact jobs your original GPT performed before switching users to the Plugin.

How to Test a Migrated GPT Plugin

The migration is not finished when the Plugin exists. It is finished when the replacement can reliably perform the jobs you depend on.

Use the Designs24hr MATCH Test to evaluate the replacement systematically.

M Method

Does the Plugin follow the same important workflow and instructions?

A Answers

Are the responses complete, accurate, structured, and useful enough for the original task?

T Tools

Do the necessary apps, integrations, file tools, or other capabilities still work?

C Context

Is it using the expected reference files and important source material correctly?

H Handoff

Can the people who need the Plugin actually install, access, and use it?

MATCH is a Designs24hr testing framework, not an official OpenAI methodology.

Run a familiar-task test

Start with something the original GPT already handled well. Use the same task with the migrated Plugin and compare the result.

Migration test prompt Copied

Complete this task using the workflow, rules, reference material, and output structure you are configured to follow. Before finishing, check that every required field, section, constraint, and source requirement has been satisfied. [Paste one task that your original Custom GPT handled reliably.]

Run an edge-case test

A simple success is not enough. Test at least one case involving ambiguity, conflicting instructions, missing information, unusual formatting, or an integration.

Hard-case validation prompt Copied

Handle this as a difficult version of your normal workflow. Follow your configured instructions, use relevant reference material, do not invent missing facts, and clearly identify anything you cannot complete with the tools or permissions currently available. [Paste a difficult or integration-dependent version of your normal task.]

Compare more than writing style

A replacement can sound good while still being functionally incomplete. Compare:

  • whether the correct workflow was followed;
  • whether important instructions were obeyed;
  • whether expected reference material was used;
  • whether the same files can be produced;
  • whether required apps and integrations are available;
  • whether the output contains all required fields;
  • whether the result is usable without extra manual repair.
Best switching rule Keep using the original GPT for an important workflow until the migrated Plugin has passed your real tests. OpenAI specifically recommends testing familiar work and at least one harder case before relying on the replacement.

What Happens to the Original GPT After Migration?

The original GPT does not disappear immediately after migration. OpenAI says it remains usable until the applicable retirement date, but it becomes read-only. Future maintenance happens in the replacement Plugin.

At retirement, Custom GPTs and their GPT pages are scheduled to become inaccessible. That is why the safest strategy is to migrate and validate the workflows you care about before the deadline rather than treating December 11 as the day to begin.

Will Existing GPT Conversations Transfer?

No. Existing GPT conversations do not move into the Plugin as part of the migration.

Before retirement, review old GPT conversations that contain information you may need later—such as useful examples, corrections, recurring preferences, approved formats, edge cases, or decisions. Preserve the important information in an appropriate location rather than assuming it will appear automatically in the Plugin.

Do not upload or preserve sensitive information merely because it appeared in an old chat. Follow your organization’s privacy, security, retention, and data-handling requirements.

What Happens to GPT Sharing and Existing Users?

Migration does not automatically preserve the audience of your original GPT. OpenAI says the replacement Plugin starts private, and existing GPT users do not automatically receive access to it.

Before announcing a switch:

  • Confirm the Plugin works for the intended workflow.
  • Review its sharing settings.
  • Confirm the intended users are eligible to access Plugins.
  • Ask at least one intended user to install and test it.
  • Update any documentation or links that still point to the Custom GPT.

For managed workspaces, sharing a Plugin and publishing it to a workspace can involve separate permissions.

Should You Migrate Every Custom GPT?

Probably not.

Migration is a good time to remove workflows that no longer provide meaningful value. If a GPT is rarely used, duplicates another tool, contains outdated instructions, or solves a task that is now easier in a normal ChatGPT conversation or Project, maintaining a Plugin may create unnecessary work.

Ask one question:

Would I rebuild this workflow today if it did not already exist? If the answer is no, migration may not be worth the maintenance. If the answer is yes, preserve the important parts, migrate carefully, and test the replacement.

If you are deciding whether an ongoing workflow belongs in a persistent workspace instead, compare the alternatives in our ChatGPT Projects vs Custom GPTs guide. Keep in mind that the Custom GPT side of that comparison is now changing because of the 2026 retirement transition.

Custom GPT Migration Checklist

Use this final checklist before considering your migration complete.

  • Identify the Custom GPTs you still genuinely need.
  • Confirm you are using the correct account or workspace.
  • Finish important GPT edits before migration.
  • Publish the version you want the migration flow to use.
  • Save important instructions and workflow notes.
  • Record expected knowledge/reference files.
  • Identify every Custom Action.
  • Save several familiar test prompts.
  • Save at least one difficult test case.
  • Use Migrate to plugin when it becomes available.
  • Review migrated instructions and reference files.
  • Verify connected apps and required permissions.
  • Replace or rebuild required Custom Actions.
  • Run the MATCH test against your original workflow.
  • Review the Plugin’s sharing settings.
  • Have an intended user test access where relevant.
  • Update old links, documentation, or onboarding material.
  • Maintain the Plugin rather than trying to edit the migrated GPT.

Frequently Asked Questions

When are Custom GPTs retiring?

OpenAI currently schedules the standard Custom GPT retirement for December 11, 2026. Certain affected Enterprise workspaces with an approved deferral can have a later date. Follow the notice shown for your own account or workspace because timelines can differ.

How do I migrate a Custom GPT to a Plugin?

When migration is available, go to My GPTs, find a GPT you created, and choose Migrate to plugin. Review the migrated instructions, reference files, apps, and tools, then test the replacement before relying on it.

Why is Migrate to plugin missing?

Migration may not be available to your account yet, you may be using the wrong account or workspace, Plugin access may be disabled, the GPT may not be eligible in its current state, or workspace permissions may prevent migration. Using someone else’s GPT does not give you permission to migrate it.

Do Custom GPT knowledge files transfer?

OpenAI says the GPT’s knowledge files are copied into the Plugin’s reference files during the planned migration flow. Review and test them after migration rather than assuming they are being used exactly as before.

Do Custom GPT instructions transfer?

Yes. Under OpenAI’s current migration workflow, the GPT’s instructions become a skill inside the replacement Plugin. Review them after migration because the Plugin can still behave differently from the original GPT.

Do Custom Actions transfer to ChatGPT Plugins?

No. OpenAI says GPT Custom Actions do not automatically transfer. Workflows that depend on them need a supported replacement, such as an available app or, where appropriate and supported, a separately built integration.

Will my old Custom GPT conversations migrate?

No. Existing GPT conversations are not moved into the replacement Plugin through the migration flow.

Will the Plugin behave exactly like my Custom GPT?

Not necessarily. The selected model does not carry over, available tools can differ, Custom Actions require separate handling, and a Plugin may follow the migrated instructions differently. Test familiar tasks and a difficult case before switching.

Can people who used my GPT automatically use the Plugin?

No. Migration does not automatically preserve your GPT’s sharing settings or grant existing users access to the replacement. Review sharing and Plugin access separately.

What happens if I do nothing before Custom GPT retirement?

At the applicable retirement date, Custom GPTs and their GPT pages are scheduled to become inaccessible. Workflows you want to keep should be migrated or otherwise replaced and tested before that date.

Official Sources and Further Reading

Product behavior can change during the transition. For the latest account-specific instructions, check your ChatGPT notice and OpenAI’s current documentation:

Final Takeaway

The safest approach to custom GPT migration is not simply clicking Migrate to plugin. First preserve what matters, confirm that your latest published GPT contains the right setup, identify anything that will not transfer, and create repeatable tests.

After migration, review the instructions and reference files, rebuild any essential Custom Actions, run familiar and difficult tasks, and verify sharing before asking other users to switch.

Treat the migration as a workflow transition rather than a file transfer. If the replacement passes your real tests, you have preserved the useful part of the GPT—not merely its name.

Independent publication notice: Designs24hr is not affiliated with or endorsed by OpenAI. ChatGPT, Custom GPTs, Plugins, and related product names are used descriptively. Features, dates, permissions, plan availability, and interface labels can change. Always follow the current notice and documentation for your own account or workspace.

Leave a Reply

Your email address will not be published. Required fields are marked *