By the end of this guide, you will know exactly what is happening to custom GPTs, what replaces them, which parts of your GPT survive the move and which parts quietly disappear, and the six things to do before the deadline. If someone in your company built a GPT that writes quotes, answers staff questions or drafts product descriptions, this affects you.
OpenAI announced on 11 September 2026 that it plans to retire custom GPTs across ChatGPT plans. The scheduled retirement date is 11 December 2026. That is roughly ten weeks from today.
What Is Happening to Custom GPTs?
In short: OpenAI is retiring custom GPTs, the build-your-own ChatGPT assistants launched in November 2023. Existing GPTs keep working until 11 December 2026, when they and their GPT pages become inaccessible. Creators can migrate each GPT to a plugin, which is OpenAI's new format for bundling instructions, reference files and connected apps.
A custom GPT is a version of ChatGPT that someone has pre-loaded with instructions and files for one job, such as "answer questions about our staff handbook" or "turn my notes into a quotation".
According to OpenAI's Custom GPT retirement and migration FAQ, the change affects all ChatGPT plans. Enterprise workspaces with an approved deferral get until 11 February 2027. Everyone else should plan around 11 December 2026.
Some doors have already closed. MIXED reported that OpenAI's GPT help page now says new GPT creation and publishing are no longer available on personal accounts (Free, Go, Plus and Pro), while existing GPTs remain usable.
What Is a Plugin, and How Is It Different From a Custom GPT?
In short: A plugin is a bundle. It can contain a skill (reusable instructions), reference files, and apps (connections to tools such as your email or files). A custom GPT was one assistant you opened on purpose. A plugin is a package ChatGPT can reach for inside ordinary chats, in Work and in Codex, when your request matches it.
Think of a custom GPT as a specialist you had to walk over to. You opened "Quote Writer GPT" and asked it for a quote. A plugin is closer to a skill your general assistant has learned. You can call it by name, but ChatGPT may also use it automatically when your request fits the skill's description.
Under OpenAI's planned migration, the pieces map across like this:
--- The GPT's instructions become a skill inside the new plugin.
--- The GPT's knowledge files are copied into the plugin's reference files.
--- The GPT's connected apps are added to the plugin as apps.
If you want the background on plugins themselves, our earlier explainer What Is a ChatGPT Plugin? covers how skills, apps and admin settings fit together.
What Carries Over to the Plugin, and What Gets Left Behind?
In short: Your written instructions, knowledge files and connected apps carry over. Your past conversations, your chosen model, any custom actions, your sharing settings and any unpublished draft edits do not. Custom actions are the biggest risk, because any GPT that talked to an outside system will stop doing that part of its job until someone rebuilds it.
OpenAI's FAQ lists five things that stay behind:
--- Conversations. Old chats with the GPT do not move into the plugin.
--- The selected model. The model you picked for the GPT does not carry over.
--- Custom actions. Connections to outside services through the GPT's Actions feature "do not transfer through the migration workflow".
--- Sharing settings. A migrated plugin starts private. Nobody who used your GPT gets automatic access.
--- Unpublished edits. Migration uses the latest published version only. Drafts are ignored.
The custom actions point deserves attention. A custom action is the part of a GPT that reaches outside ChatGPT, for example checking stock in your inventory system or creating a booking. OpenAI says rebuilding these may require a custom MCP server and technical setup. For most small businesses, that means calling whoever built the integration in the first place.
How Does the Custom GPT Retirement Affect a Small Business?
In short: The risk is not losing a chatbot. It is losing a process that quietly became part of how your business runs. A GPT that drafts quotations, answers staff policy questions or writes product listings carries your business rules. If nobody saves and tests those rules before December, the job it did simply stops.
Consider three ordinary Hong Kong scenarios.
The trading company. A sales coordinator built a GPT two years ago that turns WhatsApp enquiries into formatted English quotations using the company's price list. She has since left. The GPT still works, but nobody knows whose account it lives on, and only the creator or an admin can migrate it.
The bakery with two branches. The owner uses a public GPT made by a marketing agency to write Instagram captions. The owner did not build it, so the owner cannot migrate it. The only options are waiting for the agency's replacement plugin or recreating the instructions from scratch.
The property agency. A GPT checks listing descriptions against the agency's house style and calls a custom action to pull flat details from the office database. The writing rules will migrate. The database connection will not.
Notice that none of these is a technology problem. Each one is an ownership problem: an account nobody tracks, a tool someone else controls, or a connection nobody documented. That is good news, because ownership problems are fixed with a list and an afternoon, not a budget.
In each case, the question to ask is the same: what job does this GPT do, who owns it, and what would we do on 12 December if it were gone?
How Do You Migrate a Custom GPT to a Plugin?
In short: When migration is available on your account, go to My GPTs and select "Migrate to plugin". Before you do, publish any edits you need, save copies of the instructions and files, and write down a few typical requests and good answers. After migrating, test the plugin against those examples before telling staff to switch.
Here is a practical six-step checklist based on OpenAI's own guidance:
--- Step 1: Make a list. Write down every GPT your team relies on, who created it, which account it lives on, and the job it does.
--- Step 2: Save the business knowledge. Copy the instructions and download the knowledge files into a shared folder you control. This protects you even if migration goes wrong.
--- Step 3: Record five test cases. Save five normal requests and the answers you consider correct, plus one difficult case.
--- Step 4: Publish, then migrate. Finish essential edits and publish first (publishing does not mean making it public). Then choose "Migrate to plugin".
--- Step 5: Test side by side. Run your test cases through the plugin. Check it follows the instructions, uses the right files and produces the right format.
--- Step 6: Share deliberately. The plugin starts private. Share it with the right people and ask one colleague to install and test it.
One warning from the FAQ: after migration, the original GPT becomes read-only and its creator cannot delete it. Make every essential edit before you press the button.
What Do People Get Wrong About the Custom GPT Retirement?
In short: The three most common mistakes are assuming GPTs already stopped working, assuming migration copies everything, and assuming anyone who uses a GPT can migrate it. None of these is true. GPTs still run until December, migration leaves several things behind, and only the creator or a workspace admin can migrate one.
Misconception 1: "My GPT is already gone." It is not. OpenAI's FAQ states existing GPTs remain usable until the retirement date, subject to their existing access.
Misconception 2: "Migration is a one-click copy." It copies instructions, files and apps. It does not copy actions, chats, the selected model or sharing settings. OpenAI also warns a migrated plugin "may respond differently".
Misconception 3: "I use it every day, so I can move it." Permission to use a GPT does not give permission to migrate it. If you rely on someone else's GPT, check the notice on it and the creator's guidance.
Misconception 4: "I'm on a personal plan, so the Enterprise dates don't matter." According to the FAQ, public GPTs created in affected Enterprise workspaces follow the Enterprise transition even when you open them from a personal account. What matters is where the GPT was made.
Frequently Asked Questions About Retiring Custom GPTs
In short: Most owners ask the same five questions: when it happens, whether chats are saved, whether there is a cost, what happens to shared links, and what to do if the migrate button is missing. The answers below are drawn from OpenAI's published FAQ as of late September 2026.
When exactly do custom GPTs stop working?
The scheduled retirement is 11 December 2026. Enterprise workspaces with an approved deferral have until 11 February 2027. Follow the notice in your own account, because OpenAI says dates may differ.
Will my old GPT conversations be saved?
They will not move into the plugin. You can keep using the original GPT and its chats until retirement, so export anything important before 11 December.
Does using a plugin cost extra?
For Enterprise workspaces, OpenAI says using a migrated GPT without apps in Instant Chat has no incremental cost, while actions in Work consume credits. Check your own plan's terms for other tiers.
Why can't I see the "Migrate to plugin" option?
Migration may not have reached your account yet, plugins may be disabled by your admin, or you may be signed in to a different workspace from the one where the GPT was created.
What should I do this week?
Complete Step 1 and Step 2 of the checklist above. A list of your GPTs plus saved copies of their instructions takes under an hour and removes most of the risk.
Conclusion: Save the Process, Not Just the Chatbot
In short: Custom GPTs retire on 11 December 2026. The instructions, files and apps can move to plugins, but conversations, custom actions, model choice and sharing settings do not. The businesses that come through this smoothly will be the ones that treat each GPT as a documented process, not a gadget.
The deeper lesson is one that will repeat. AI tools change shape every few months. What lasts is the business knowledge you put into them: your price rules, your tone, your standard answers. Keep that knowledge somewhere you own, and moving it into the next tool becomes a one-afternoon job instead of a crisis.
That is exactly the kind of change where a steady partner helps. We understand AI. UD stands with you.
Reviewed by the UD AI team.
Before your GPTs disappear, find out which parts of your business are ready for AI and which processes need writing down first. Take the free AI Ready Check, and we will walk you through it step by step, from your first result to a plan your team can actually follow.