Custom GPTs are going away: what to use instead

Short answer: OpenAI is retiring custom GPTs on December 11, 2026. They are being replaced by plugins. Your GPT’s instructions and knowledge files can move across, but the model you chose, any custom Actions and your sharing settings do not come with them. Your past conversations stay accessible. Start by listing every GPT your team uses, then export its instructions and files now, so nothing depends on a feature that is about to disappear. Enterprise workspaces with an approved deferral have until February 11, 2027.

This guide is for people who built or rely on custom GPTs and are not developers. If you are not sure what a custom GPT is, start with the custom GPT glossary entry.

What is changing, and when

• December 11, 2026: custom GPTs retire.
• February 11, 2027: the end date for Enterprise workspaces that have an approved deferral. If your company is on Enterprise, ask your admin whether a deferral was approved for you.
• After that: a custom GPT is no longer something you can open and use. Anything you have not exported or migrated is at risk.

OpenAI’s replacement is called a plugin. A plugin bundles the same kinds of things a GPT held: reusable instructions, reference files and, in some cases, connected apps.

What carries over, and what does not

In your custom GPT	What happens when it migrates
Instructions	Become a Skill inside the plugin
Knowledge files	Become reference files
Connected apps	May be included in the plugin
Existing conversations	Stay accessible after retirement, with no action needed
The model you picked	Not kept
Custom Actions	Do not transfer automatically. They have to be rebuilt, through an app or an MCP server
Sharing settings	Do not carry over. A migrated plugin starts out private

Two of these surprise people. First, a GPT that called an outside service through an Action will not work after migration until someone rebuilds that connection. If you are not sure whether your GPT has an Action, open its settings and look for one. Second, a GPT you shared with the whole team becomes private, so you have to share the new plugin again.

If you need a refresher on what MCP is, it is a standard way to connect an AI tool to another app or data source.

A checklist to finish before December 11

Work through this in order.

1. Inventory your GPTs. List every custom GPT you built and every one your team uses, including ones built by colleagues who have left. Write down the name, the owner and the link.
2. Note who uses each one, and how often. A GPT that three people open daily needs migrating first. One that nobody has opened since spring can probably be retired.
3. Copy the instructions out. Open each GPT, copy the full instructions into a plain document and name the file after the GPT. This is the most useful piece, and it takes five minutes each.
4. Download the knowledge files. Save every file you uploaded, and note which ones are out of date so you do not carry stale material across.
5. Check for Actions and connected apps. Mark every GPT that calls an outside service. These need someone technical, or a rebuilt connection, after the move.
6. Write down the model and the sharing. Record which model you chose and who the GPT is shared with, since neither migrates.
7. Decide: migrate, move, or retire. Migrate to a plugin if the team lives in ChatGPT. Move to another tool if you want to stop depending on one provider. Retire anything unused.
8. Test the result with three real requests that people actually make, and compare the answers to what the old GPT gave.
9. Share the new version and tell the team. Migrated plugins start private, so re-share them, and say where the old GPT’s replacement lives.
10. Keep the instructions in a document you own. A plain document works in any tool. This is what makes the next change a ten-minute job.

What to use instead

OpenAI plugins. The default path if your team already works in ChatGPT. Migration is built in, so instructions and files carry over. You re-check sharing, the model and anything that used an Action.

Claude Projects. Anthropic’s Claude has Projects, which hold instructions and files for repeated work. Pasting your exported instructions into a Project and uploading the same files rebuilds the basic behavior of a GPT, with no coding.

Gemini Gems. Google offers Gems in Gemini, which work the same way: saved instructions you reuse. Same approach, paste in the instructions and add the files. Google’s help page says Gems on personal Google accounts will transition to skills starting in November 2026, so check what applies to your account before you build there.

Feature names and plan limits in all three products change often. Check each product’s current help pages before you commit your team to one. For help choosing, see which AI tool should I use at work and Claude vs ChatGPT.

Make the instructions portable

The reason migrations hurt is that the instructions only existed inside one product. Write them so they work anywhere:

• State the one job the assistant does.
• List the steps in order, and the format of the output.
• Say what to do when information is missing (“ask me, do not guess”).
• Keep reference files short, current and approved, since the assistant quotes them.

How to write AI prompts that work covers the writing in detail, and system prompts explains what the instructions in a GPT actually are.

What OpenAI’s FAQ says

Each line below is taken from OpenAI’s own migration FAQ. Check it again for your account before the deadline.

• OpenAI says custom GPTs are scheduled to retire on Dec 11, 2026.
• OpenAI says FedRAMP and affected Enterprise workspaces with a qualifying, approved deferral have a retirement date of Feb 11, 2027, and that this later date does not apply to every workspace.
• OpenAI says that during migration, the GPT’s instructions become a skill within the new plugin.
• OpenAI says the GPT’s selected model does not carry over, and that in Enterprise workspaces, Enterprise defaults apply.
• The OpenAI Help Center says GPT custom actions do not transfer through the migration workflow.
• OpenAI says a migrated plugin starts private, and migration does not automatically give other people access.
• OpenAI says migration uses the GPT’s latest published version, and drafts and unpublished edits do not transfer.
• OpenAI says existing conversations with custom GPTs will remain accessible after retirement.
• OpenAI says the original GPT remains usable until retirement but becomes read-only after migration.
• Anthropic says Projects are self-contained workspaces with their own chat histories and knowledge bases, and that free Claude accounts can create a maximum of five projects.
• Google says that starting in November 2026, Gems will go away on personal Google accounts and automatically transition to skills.
• In Gemini Apps Help, Google says you do not need to use all four areas of Gem instructions, but using a few will help.

Common questions

Will I lose my GPT’s chat history? No. OpenAI says existing conversations with custom GPTs remain accessible after retirement and you do not need to act to preserve them.

Can I keep using my GPT until the last day? According to OpenAI’s migration FAQ, December 11, 2026 is the retirement date. Do the export steps before then so that a last-minute change does not catch your team out.

Do I have to use OpenAI’s plugins? No. Your instructions and files are plain text and documents, so they can be rebuilt in Claude Projects, Gemini Gems or another tool.

What if my GPT used an Action? It needs rebuilding as an app connection or through an MCP server. That is usually a job for someone technical on your team, or for a short working session.

Moving a handful of GPTs is the kind of thing that goes quickly when you have the checklist and someone to sit next to. Learn AI Fast is a two-day in-person class in Boston for non-technical professionals where you practice building AI-assisted workflows.

Sources

• OpenAI Help Center, Custom GPT retirement and migration FAQ: “Custom GPTs are scheduled to retire on Dec 11, 2026”
• OpenAI Help Center, Custom GPT retirement and migration FAQ: “a qualifying, approved deferral have a retirement date of Feb 11, 2027. This later date does not apply to every workspace”
• OpenAI Help Center, Custom GPT retirement and migration FAQ: “instructions become a skill within the new plugin”
• OpenAI Help Center, Custom GPT retirement and migration FAQ: “selected model does not carry over. In Enterprise workspaces, Enterprise defaults apply”
• OpenAI Help Center, Custom GPT retirement and migration FAQ: “GPT custom actions do not transfer through the migration workflow”
• OpenAI Help Center, Custom GPT retirement and migration FAQ: “The plugin starts private, and migration does not automatically install it or give other people access”
• OpenAI Help Center, Custom GPT retirement and migration FAQ: “Migration uses the GPT’s latest published version. Drafts and unpublished edits do not transfer”
• OpenAI Help Center, Custom GPT retirement and migration FAQ: “will remain accessible after retirement”
• OpenAI Help Center, Custom GPT retirement and migration FAQ: “becomes read-only after migration”
• Anthropic Help Center, What are projects?: “Projects allow you to create self-contained workspaces with their own chat histories and knowledge bases”
• Gemini Apps Help, Use Gems in Gemini Apps: “Starting in November 2026, Gems will go away on personal Google accounts and automatically transition to skills”
• Gemini Apps Help, Tips for creating custom Gems: “You don’t need to use all four, but using a few will help”