AI Academy Blog

Goodbye Custom GPTs. Hello Plugins. What ChatGPT Users Need to Know Before December 11

Written by Jeremy Zimmer | Nov 20, 2026, 8:16:44 PM

OpenAI is retiring custom GPTs on December 11, 2026. Here's what happens to the GPTs you've built, how Plugins compare, and what you can do now to preserve your AI workflows.

If you've spent time building custom GPTs in ChatGPT, there's an important deadline approaching.

On December 11, 2026, OpenAI is scheduled to retire custom GPTs, replacing the familiar experience with newer ways to build and use reusable AI capabilities.

Maybe you created a GPT that helps you brainstorm marketing campaigns. Maybe your team uses one to analyze research, generate reports, or answer questions using company knowledge. Or maybe you've come to rely on GPTs created by other people.

Whatever your use case, the feature sunset raises an important question: What happens to the time, knowledge, and effort you've invested in these tools?

The good news is that you don't necessarily have to start over.

In our new AI Academy course, ChatGPT Update: Goodbye GPTs. Hello Plugins., I explore what's changing, what alternatives are available, and how to preserve much of the functionality you've already built.

Here's a recap of what you need to know.

What's happening to custom GPTs?

OpenAI has announced that custom GPTs will be retired on December 11, 2026, with later dates applying to certain eligible workspaces.

For most users, that means the GPTs they've created or regularly use will no longer be accessible after the retirement date.

However, there are some important distinctions.

Your existing conversations aren't going away. OpenAI has confirmed that conversations with custom GPTs will remain accessible in your ChatGPT history.

Your GPT configurations are another story. Once retirement takes effect, the GPTs themselves—including their configuration pages, instructions, and associated knowledge files—will become inaccessible.

That makes it important to preserve the underlying materials before the deadline.

And if you rely on GPTs created by others, you'll need to check whether their creators plan to offer replacements. Access to an existing GPT doesn't automatically carry over to a new Plugin. More FAQs listed by OpenAI here.

The takeaway? You don't have to abandon the workflows you've developed, but you do need a plan to preserve them.

What replaces custom GPTs?

OpenAI is introducing several ways to organize and extend AI capabilities in ChatGPT. Taylor's course explores three alternatives: Plugins, Projects, and Sites.

Each serves a slightly different purpose.

Plugins: Reusable AI capabilities

Plugins are OpenAI's recommended replacement for many custom GPT use cases.

They allow you to package specialized instructions, reference materials, and connected apps into reusable AI tools that can be accessed within ChatGPT.

If you built custom GPTs to follow consistent instructions, produce repeatable outputs, or apply specialized knowledge, Plugins are likely the closest alternative.

Projects: Shared context for ongoing work

Projects bring related conversations, files, and instructions together in one place.

If you primarily used a custom GPT to maintain context for a specific initiative, responsibility, or ongoing project, you may find that a Project meets your needs without requiring a Plugin.

For example, you might use a Project to organize campaign planning, research, or an ongoing business initiative.

Sites: Custom experiences and applications

Sites offer another approach: building websites or lightweight applications with ChatGPT.

If your goal is to provide a more customized, interactive experience, Sites may be worth exploring.

However, this generally involves rebuilding the experience rather than directly migrating an existing GPT.

The best alternative depends on what you're trying to accomplish—not simply which feature replaces the one you're losing.

How do Plugins compare to custom GPTs?

At first glance, Plugins and custom GPTs have a lot in common.

Both allow you to create specialized AI experiences, provide instructions, and incorporate supporting knowledge.

The difference is how those capabilities are organized.

With a custom GPT, you configured the instructions, knowledge files, and additional capabilities together inside a single assistant.

Plugins take a more modular approach.

A Plugin can contain a Skill, which defines the instructions and steps ChatGPT should follow. It can also contain reference files, including documents and supporting information, and connected apps that provide access to external services or actions.

When you migrate a GPT into a Plugin, OpenAI can convert your existing instructions into a Skill and copy your knowledge files into the Plugin's reference files.

That means much of the work you've already invested can be repurposed.

There's an important exception, though: Custom actions don't automatically transfer. If your GPT depends on specialized integrations or external services, you may need to rebuild those connections separately.

A migrated Plugin also may not behave exactly like the original GPT, even when the instructions are the same. Testing is essential.

Two ways to migrate an existing GPT into a Plugin

In the course, we demonstrate two approaches to transitioning an existing custom GPT.

Option 1: Use OpenAI's built-in migration feature

The simplest approach is OpenAI's migration feature.

When the option is available in your ChatGPT account, you can navigate to My GPTs, select a GPT you've created, and choose Migrate to plugin.

ChatGPT then converts the GPT's published instructions into a Skill and transfers its knowledge files into the new Plugin.

Taylor demonstrates this process using an English-to-Japanese translation GPT she previously created.

The process is relatively straightforward, and the resulting Plugin can be reviewed, tested, and refined.

However, there's an important trade-off.

Once you migrate a GPT using this feature, the original GPT becomes read-only.

You can continue using it until retirement, but you can no longer edit its configuration. Future changes must be made to the Plugin.

Before choosing this option, make sure you've finalized and published any important changes to your existing GPT.

Option 2: Recreate your GPT using Plugin Creator

The second approach gives you more flexibility.

Instead of using the built-in migration feature, you can manually recreate an existing GPT using Plugin Creator, a tool available within ChatGPT.

Despite the name, the process doesn't require you to write code or build everything from scratch.

You can provide your GPT's original instructions, upload its knowledge files, and ask Plugin Creator to build a new Plugin using those materials.

Taylor demonstrates this approach using a personal GPT called Cube Chef, which helps turn prepared freezer ingredients into meal ideas.

She copies the original instructions, uploads a recipe guide, and uses Plugin Creator to build a replacement.

The advantage is that the original GPT remains editable throughout the process.

You can compare the original GPT and the new Plugin side by side, test them using the same prompts, and adjust the Plugin until you're satisfied with its performance.

If you rely heavily on a particular GPT—or want time to evaluate the replacement before switching—this approach may be worth considering.

What about using and sharing Plugins?

Once you've created or installed a Plugin, the experience is similar to using a custom GPT.

You can browse available Plugins in ChatGPT, install those relevant to your work, and invoke them in conversations using an @ mention or other supported selection options.

But one of the biggest changes involves sharing.

Previously, custom GPT creators could make their tools available to anyone with a link or publish them to the GPT Store.

Plugins work differently.

New Plugins are private by default. Within supported ChatGPT Business and Enterprise workspaces, creators can share Plugins with specific colleagues, teams, or other eligible workspace members, depending on permissions.

Public distribution requires a separate submission and approval process through OpenAI.

For organizations that created GPTs primarily for internal use, workspace sharing may provide a practical solution.

But for creators who built publicly accessible GPTs, this is a significant consideration.

You can't assume that an existing GPT's users will automatically have access to its replacement—or that an approved public Plugin will be available immediately.

Taylor walks through these sharing options in the course, including how they differ between personal and business accounts.

What should you do before December 11?

If you've built or regularly use custom GPTs, now is the time to evaluate which ones are worth preserving.

Taylor recommends taking several practical steps:

  1. Review your existing GPTs. Identify which ones you use regularly, which support important workflows, and which you've shared with colleagues or customers.
  2. Preserve your instructions and knowledge files. Save copies of the materials that make each GPT useful before access to its configuration disappears.
  3. Determine the right replacement. Consider whether a Plugin, Project, or another approach best supports the original purpose of your GPT.
  4. Choose a migration approach. Decide whether to use OpenAI's built-in migration feature or manually recreate the experience using Plugin Creator.
  5. Test before switching. Run familiar prompts through the replacement and evaluate whether it follows your instructions, references the correct materials, and produces the outputs you expect. Confirm that intended users can access it.

If you use GPTs created by someone else, check whether the creator is planning a replacement and how you'll be able to access it.

For more information, review OpenAI's custom GPT retirement and migration FAQ and Plugins documentation.

The bigger lesson: Build workflows that can evolve

The retirement of custom GPTs illustrates a broader reality of working with artificial intelligence: The technology will keep changing.

New models, capabilities, products, and interfaces are being introduced rapidly. Features that feel foundational today may eventually be replaced by something more capable—or simply different.

That doesn't mean the work you've invested in developing effective AI workflows loses its value.

The instructions you've refined, the knowledge you've organized, and the processes you've developed to solve real problems can remain useful even when the tools change.

The interface may change, but the methodology shouldn't.

Rather than thinking about your AI workflows as belonging to a particular product or feature, think about the problem you're solving and the capabilities required to solve it.

That perspective makes it easier to evaluate new tools, adapt existing workflows, and continue building on what you've learned.

Learn how to migrate your GPTs in AI Academy

Want to see the transition in action?

In ChatGPT Update: Goodbye GPTs. Hello Plugins., Taylor Radey walks through the changes in ChatGPT and demonstrates how to migrate an existing GPT, manually recreate one using Plugin Creator, and test and share the resulting Plugins.

The course also explores the advantages and limitations of the new approach so you can make informed decisions about the AI tools and workflows you rely on.

Whether you've built one custom GPT or manage several across your organization, you'll leave with a clearer understanding of what to preserve, what to reconsider, and what to do next.

Watch ChatGPT Update: Goodbye GPTs. Hello Plugins. in AI Academy by SmarterX.

Access the Course in AI Academy