Gareth MorgansProduct
Custom GPTs are becoming ChatGPT plugins. Plan the move.
ChatGPT plugins are taking over from custom GPTs. OpenAI plans to retire custom GPTs and help the people who built them move their work into plugins, with a migration tool to make the switch.
For a team that relies on a handful of GPTs, that is a real deadline. The most valuable part of each GPT, its instructions, is also the easiest part to keep.
Start free beta · Skills registry
What is changing, and when
According to OpenAI’s custom GPT retirement and migration FAQ, custom GPTs are scheduled to retire on 11 December 2026, and the transition affects all ChatGPT plans. On the retirement date, custom GPTs and their GPT pages become inaccessible. Until then, existing GPTs keep working.
OpenAI also lists planned milestones for affected Enterprise workspaces. In that list, creation of new custom GPTs was planned to end on 26 October 2026. OpenAI advised publishing any drafts you want to migrate before that cutoff, and says the freeze does not stop edits to existing GPTs that have not been migrated. Enterprise workspaces with a qualified, approved deferral retire on 11 February 2027 instead, and that extension does not apply to every workspace. OpenAI notes that these dates are subject to change, so check the notice in your own account or workspace.
Personal accounts, including Free, Go, Plus and Pro, already cannot create or publish new GPTs, according to OpenAI’s guide to sharing and publishing GPTs. Past conversations are safe. OpenAI says your existing conversations with custom GPTs will remain accessible after retirement.
What ChatGPT plugins are
A ChatGPT plugin brings together the guidance and tools a task needs. OpenAI’s help article on plugins describes four parts: skills, which provide instructions and workflow guidance; connected apps, which link external accounts and services; app templates, which help administrators set up apps for a workspace; and extensions, which add interactive views, forms and settings.
Skills and apps matter most here. A skill holds reusable instructions, such as a report format or a house style. An app connects ChatGPT to a service such as Slack or Google Drive. Some plugins contain only skills and need no connection at all.
The plugin directory is available across ChatGPT plans, although what you can install depends on your plan and workspace. Skills in ChatGPT are available to eligible Business, Enterprise, Healthcare and Edu users, and members of Business and Enterprise workspaces can share plugins with each other.
What moves across, and what does not
OpenAI’s planned migration turns each GPT into a plugin. The GPT’s instructions become a skill within the new plugin. Its knowledge files are copied into the plugin’s reference files. Connected apps are added to the plugin as apps.
Some things stay behind. Custom actions do not transfer, so anything that depends on them will not work until you set up a replacement, which may mean an available app or a custom MCP server. The GPT’s selected model does not carry over. Nor do its sharing settings, so the people who used the GPT do not get access to the plugin automatically.
Migration uses the GPT’s latest published version, so drafts and unpublished edits are left out. The new plugin starts private. The original GPT becomes read-only, and while it stays usable until retirement, every future change belongs in the plugin.
OpenAI also cautions that a migrated plugin may respond differently. Test it with familiar prompts and at least one harder case before anyone relies on it.
Your instructions are the asset
Look at what needs rework in that list: the model, the sharing and the integrations. The instructions move across as a skill. They are the part your team spent months refining, and the part worth protecting.
Instructions carry how your team writes, what it checks and what it never says. They are also the part most likely to exist in only one place, typed into a GPT editor and quietly edited there ever since.
Tools will keep changing. Instructions that live inside one tool have to be rebuilt each time they do.
Keep one source outside any single tool
The safer habit is to keep the master copy of your instructions somewhere you control, and to treat every GPT, plugin or project as a copy of it. If the idea is new, here is what a skills registry is.
That is the job of a skills registry. In Joyst, each skill has a version history and a published version, and you can roll back if a change makes answers worse. Permissions decide who can view, use or manage each skill, and the people with manage rights own it. Our guide to how you version and share skills explains the detail.
For coding tools, Joyst provides translations of each skill for Claude Code, Cursor, Codex CLI, Copilot and Windsurf. That is the idea behind skill portability, and it is why teams manage skills across AI tools from one place. Teams can apply the same approach to ChatGPT custom instructions.
Joyst does not migrate GPTs, import or export ChatGPT plugins, or connect to ChatGPT accounts. When you build or update a ChatGPT skill, you copy the instructions from Joyst by hand. Everyone copies from the same, current source.
A short plan before 11 December
OpenAI’s own advice points the same way as good housekeeping. Five steps cover most teams.
List what you rely on. Note every GPT your team uses, who created it and whether it uses custom actions. GPTs with custom actions need the most time.
Name an owner. Each GPT needs one person responsible for its replacement. Our post on who owns each skill covers how to choose.
Publish final edits. Migration uses the latest published version, and the original becomes read-only afterwards. Finish essential changes first. Publishing does not require sharing the GPT publicly.
Migrate and test. When migration is available, select Migrate to plugin in My GPTs, then compare the plugin’s answers with the GPT’s on the same prompts. Rebuild any custom actions and confirm who has access before asking people to switch.
Keep the master copy in your registry. Save each set of instructions in one versioned place, make future changes there first, and update the ChatGPT skill from it.
FAQ
Are custom GPTs going away?
Yes, under OpenAI’s current plan. Custom GPTs are scheduled to retire on 11 December 2026. Enterprise workspaces with an approved deferral have until 11 February 2027. OpenAI says the dates are subject to change.
Do my old GPT conversations disappear?
No. OpenAI says existing conversations with custom GPTs will remain accessible after retirement, and you do not need to do anything to keep them.
Do custom actions transfer to ChatGPT plugins?
No. Custom actions do not transfer through the migration workflow. You will need to set up a replacement, such as an available app or a custom MCP server, and test it before switching.
Does Joyst migrate GPTs or publish ChatGPT plugins?
No. Joyst does not migrate GPTs, import or export ChatGPT plugins, or publish to the plugin directory. It holds the versioned source of your instructions, which your team copies into ChatGPT by hand.