Custom GPTs in ChatGPT are approaching retirement. Learn what can still be edited, how instructions, knowledge, capabilities, and Preview work, and how creators should prepare for migration.
What custom GPTs in ChatGPT are—and their current status
Custom GPTs in ChatGPT combine reusable instructions, uploaded reference files, and optional tools in a configured ChatGPT experience. However, OpenAI is preparing to retire them and move creator workflows toward plugins (OpenAI Help Center).
New creation and publishing are unavailable on personal Free, Go, Plus, and Pro accounts. Availability in managed workspaces depends on workspace controls and permissions. OpenAI lists October 26, 2026, as the end of creation for affected Enterprise workspaces other than FedRAMP workspaces. Existing, unmigrated GPTs may still be used and edited until their applicable retirement date.
For someone planning a reusable workflow today, the practical recommendation is to avoid investing in a new custom GPT unless the relevant workspace still permits creation and there is a clear short-term need. Existing creators should focus on reviewing, documenting, and preparing their GPTs for migration.

Instructions, knowledge, and capabilities
The configuration separates several concerns. Instructions govern behavior, response style, and boundaries across conversations. Knowledge supplies reference material from uploaded files; OpenAI documents a limit of 20 files per GPT and 512 MB per file. Capabilities can add functions such as web search, image generation, Canvas, data analysis, or connected apps, subject to account, workspace, and regional availability (OpenAI Help Center). Actions instead connect APIs defined by the creator, and a GPT cannot use actions and apps simultaneously.
As an editorial recommendation, put operating rules in instructions and reserve knowledge for material the GPT should consult. Write instructions as concrete steps, add short examples where classifications or output formats matter, and prefer text-forward reference files. Enable only the capabilities needed for the stated task; this keeps testing focused and makes unexpected behavior easier to trace.

Preview testing for an existing GPT
OpenAI’s editor includes Preview for checking prompts, tone, and accuracy before changes are saved or applied (OpenAI Help Center).
A suggested test set should include:
- A routine prompt representing the GPT’s main purpose.
- An ambiguous request that should lead to clarification.
- A request missing essential source information.
- A difficult example involving the required format or a tool.
- A prompt that falls outside the intended scope.
Record the prompts and desired characteristics before revising anything. Change one configuration area at a time, repeat the same prompts, and compare the results. This is a suggested evaluation method, not a guarantee of consistent future responses. Before enabling another tool, first consider tightening the instructions or adding a concise example.
Preparing an existing GPT for migration
OpenAI schedules standard retirement for December 11, 2026; FedRAMP workspaces and qualifying Enterprise workspaces with an approved deferral have a February 11, 2027 date. Account-specific notices control the applicable timeline (OpenAI migration guidance).
When migration becomes available, an eligible creator can open My GPTs, choose the GPT, and select Migrate to plugin. Migration uses the latest published version rather than an unpublished draft. Instructions become a plugin skill, knowledge files become reference files, and connected apps are added. The selected model and custom actions do not transfer. After migration, the original GPT remains usable until retirement but becomes read-only.
Suggested preparation:
- Finish essential edits and publish the version intended for migration.
- Inventory instructions, files, apps, actions, owners, and intended users.
- Save representative prompts and expected output requirements.
- Plan separate technical work for workflows that depend on custom actions.
- Test the replacement before directing users to it.
A migrated plugin may behave differently, so migration should not be treated as proof of equivalent results.
Conclusion
Custom GPTs in ChatGPT remain relevant mainly as existing workflows that need maintenance and transition planning. Creators should review their account notice, complete important edits, preserve a compact test set, and migrate when the option becomes available. Anyone who only uses another person’s GPT should look for guidance from its creator; access to the original does not automatically provide access to a replacement (OpenAI migration guidance).
The practical recommendation is to treat Preview testing as preparation for transition: establish the current GPT’s expected behavior, then use the same prompts to evaluate its replacement. Prioritize workflows with actions or specialized access because they may require additional setup and permission review.
Sources
- Creating and editing GPTs | OpenAI Help Center
- Custom GPT retirement and migration FAQ | OpenAI Help Center
Disclosure
This article was prepared with AI assistance.
Improve this guide
Report a problem or suggest the next question
Submissions enter a private editorial queue. They are reviewed and deduplicated before any public update or new page is created.
