Removing Custom GPT Creation from Plus and Pro Is a Bad decision

ChatGPT Feature requests
1

I’m extremely disappointed with OpenAI’s decision to remove the ability to create and publish new custom GPTs from personal ChatGPT accounts, including Plus and Pro, while keeping GPT creation available for Business, Enterprise, and Edu workspaces.

As a ChatGPT Plus subscriber, I find this change especially frustrating. Custom GPTs were one of the most useful features of the subscription, and many people use them individually for professional work.

Not everyone who uses custom GPTs is part of a company or educational institution. Freelancers, consultants, developers, creators, entrepreneurs, and other independent professionals may need custom GPTs for their own workflows without needing a Business workspace.

Removing an existing capability from paying individual subscribers and then making it available only through organizational plans gives the impression that this decision is primarily about pushing users toward more expensive plans rather than improving the product.

If a Plus or Pro subscriber is already paying for ChatGPT and only needs custom GPTs for individual use, why should they have to move to a Business plan just to regain a feature that was previously available to them?

This feels unnecessarily restrictive and, frankly, greedy.

I strongly urge OpenAI to reconsider this decision and restore the ability to create and publish custom GPTs for Plus and Pro subscribers.

I’d also be interested to hear how other Plus and Pro users feel about this change.

2

this is the letter (email ive sent to then as they are on shaky legal ground for uk law)
I was recently given less than 8 hours, post decision, that the ability to make custom publicly available gpt had been rescinded. This is not the first time that a future had been removed or “updated” which significantly changed how the product worked.
This has caused me problems in the past and ive contacted you previously about these unannounced and with immediate effect changes.

This latest change is a real material downgrade of the paid product, not a minor tweak!
This particular change has a direct and significant impact on me. I have been developing multiple projects over the last 8 months specifically intended for public release once they reached a sufficiently mature state. I withheld publishing them deliberately until they were ready. That point was reached only recently, and I am now unable to publish work I have spent 8 months developing, due to a policy change I was given no advance notice of.

I need clarification on the following before I can determine next steps:

  1. My GPTs were created before this change took effect. Does the current restriction block publishing of GPTs that already existed prior to the notification date, or does it apply only to GPTs created from that date onward? If the former, on what basis are pre-existing drafts being retroactively blocked from publishing?

  2. If publishing of pre-existing drafts is genuinely blocked, is any transition period, grace period, or exception process available for users with GPTs already in development before this change?

Given the lack of notice and the direct impact on work already completed, a restoration until my current billing cycle is the minimum remedy I consider appropriate (including the days i’ve missed since), but it does not address the substantive harm of being unable to publish 8 months of completed work. I would like OpenAI to address the publishing question above as the priority, and to confirm the restoration of FULL advertised service.

If this is not resolved satisfactorily, I will consider a card chargeback on the basis that the service is no longer as described, and will raise this with Citizens Advice / relevant consumer protection channels given the repeated pattern of undisclosed changes.

I would appreciate a substantive response addressing both points above, not an automated reply.

I would encourage others affected to send a similar email.

3

I am extremely disappointed by the announcement I received indicating that, effective today, August 22, 2026, OpenAI is removing the ability to create Custom GPTs from the ChatGPT Plus plan.

I have been a ChatGPT user since the platform first became available, and Custom GPTs have been one of the most valuable and useful features for my professional work. I also incorporate them into the Artificial Intelligence courses I teach, including courses on the application of AI in real estate appraisal and on integrating Artificial Intelligence into personal and professional life.

This is a very disappointing decision, particularly for long-time users who have invested significant time learning, developing, and teaching workflows based on capabilities that were included in the Plus subscription.

Going forward, I will recommend that my students consider and become familiar with other AI platforms in addition to ChatGPT. Changes of this nature make it difficult for professionals, educators, and businesses to rely on a platform when important capabilities can be removed from a subscription plan with limited continuity for existing users.

I sincerely hope OpenAI reconsiders this decision or provides Plus users with a practical alternative that preserves the functionality and flexibility that Custom GPTs have provided.

4

I agree with you. I also a subscriber and this change should have been at least be announced to prepare.

It is very discriminatory based on status. o just graduated so now goodbye to customgpt to subtscriber for what reason???

5

I agree with this thread, and I want to add a concrete example of why this policy is frustrating for individual builders.

I’m a Plus user and I had just finished building a new Custom GPT from scratch: system instructions, a structured knowledge base, an item bank, scoring logic, adaptive flow, output rules, evaluation criteria, branding, and pilot testing.

The GPT itself can still be created and used privately, but under the new policy I can no longer publish it to the GPT Store or even share it via link from a personal account.

That changes the value of the Builder quite dramatically.

For independent creators, freelancers, researchers, consultants, or simply advanced users, Custom GPTs were not just a toy feature. They were one of the easiest ways to move from “using AI” to actually designing and testing AI products without needing a full API deployment or a company workspace.

Requiring a Business workspace with multiple seats just to distribute a GPT is disproportionate for solo builders.

I understand that OpenAI may be trying to reduce spam, low-quality GPTs, moderation costs, or marketplace clutter. But there are much better middle-ground options:

  • limited publishing for Plus/Pro;

  • link sharing without Store access;

  • verified individual builder accounts;

  • publishing quotas;

  • review gates;

  • a dedicated individual Builder plan.

Any of these would preserve quality control without effectively removing distribution from personal builders.

The frustrating part is not that private creation is gone — it isn’t. The frustrating part is that the entire distribution layer has been removed.

I hope OpenAI reconsiders this. Independent builders are exactly the kind of users who experiment, prototype, create niche tools, and help expand the ecosystem. Locking distribution behind organization-level plans seems like the wrong tradeoff.

6

I agree.

I am a long-term paying ChatGPT user, and the ability to create and share Custom GPTs has been one of the primary reasons I have chosen ChatGPT over competing AI platforms.

I work in education and use Custom GPTs for, among other things, individually adapted compensatory support for students.

This is not simply a matter of convenience.

For some students, the ability to use a carefully configured GPT means that the student does not need to understand prompt engineering, repeatedly explain their needs, or adapt to a general-purpose AI assistant. The pedagogical structure, language level, instructions, scaffolding and working methods can instead be built into the GPT.

Different students can therefore receive different tools based on their individual educational needs.

The students use these GPTs through their own ChatGPT accounts, with their own conversations. In several cases this has also resulted in them obtaining paid ChatGPT subscriptions.

I estimate that my work and recommendations have directly or indirectly generated approximately 50-100 paying ChatGPT accounts.

The recent decision to remove the ability for personal Plus and Pro users to create and share new Custom GPTs fundamentally changes this.

I do not need to publish these GPTs in the GPT Store.

What I need is the previous functionality:

Create Custom GPT → Anyone with the link → Can chat

An unlisted link is sufficient.

Moving GPT creation to Business, Enterprise and Edu does not solve this use case. I am the creator, while the people using the GPTs are independent ChatGPT users. They are not members of one corporate workspace.

If this functionality is no longer available in ChatGPT, I will have to recreate these tools on another platform such as Gemini or Microsoft Copilot.

The consequence is therefore not only that I have less reason to maintain my own ChatGPT subscription. Future students and other users whom I introduce to these tools will naturally be introduced to the competing platform on which I can actually distribute them.

I strongly encourage OpenAI to distinguish between:

  1. Creating a Custom GPT

  2. Sharing an unlisted GPT by link

  3. Publishing a GPT publicly in the GPT Store

These are different use cases and do not need to have the same account requirements.

Restricting public GPT Store publishing to organizational plans would not prevent my educational use case.

Removing private/unlisted link sharing does.

I would therefore strongly encourage OpenAI to restore Custom GPT creation and unlisted link sharing for Plus and Pro subscribers, even if public GPT Store publishing remains restricted.

For me, this functionality is important enough that its removal affects my decision whether to continue using ChatGPT as my primary AI platform.

7

I have been thinking about this topic, and I agree that this change was a bit abrupt, with only a few days to adjust the rollout of new services via custom GPTs. This is valuable feedback.

Have you seen the announcement about ChatGPT Sites with WebMCP support?

It is a different process for creating and maintaining a service for a limited number of users with custom resources and actions, but it does check all the boxes if the user is accessing it through the ChatGPT desktop app.

What’s your opinion on this solution as the path forward?

8

I would like to add another concrete example from Japan.

I am an individual ChatGPT Plus user, not a professional software developer. I have been developing an adaptive diagnostic tool for understanding how people work with generative AI.

The diagnostic itself is already working as a Custom GPT. It dynamically generates questions, observes differences in user behavior across situations, determines when enough information has been collected, and then produces an individualized analysis.

My next step was very small: I simply wanted several members of a local AI community to try it and give me feedback.

Under the previous system, “Anyone with the link” would have been enough.

After discovering that I could no longer share a newly created GPT from my personal account, I tried using a shared normal ChatGPT conversation as a workaround.

Technically, it worked — but only because I had to put the complete instructions and diagnostic definitions into the conversation itself. When the conversation was shared, those internal instructions were visible to the participant.

So this is not a practical replacement for Custom GPT link sharing.

The remaining path is to rebuild the tool using Apps SDK/API or another application platform.

That is technically possible, but it changes the nature of the problem completely.

I started with an idea for a diagnostic tool. I was able to design it, test it, and make it work with ChatGPT without being a professional developer. But now, simply allowing a few other people to use what I built requires moving into application development.

I think this is the important issue for individual creators.

The loss is not only “GPT Store publishing.” It is the loss of the simple bridge:

Idea → Build with AI → Test → Share with a few people.

Please consider restoring unlisted link sharing for individual Plus/Pro users, even if public GPT Store publishing remains restricted.

That small capability makes a very large difference to educators, researchers, creators, community organizers, and other individual builders.

9

I agree that removing the Custom GPT builder and sharing feature impacts a lot of what I do in ChatGPT. I love the Custom GPTs and the GPT Store. OpenAI has clearly moved away from supporting it as much as they used to first by hiding the feature deep in the menus and now we can’t share them with the pubic… I am so glad I’ve have Custom GPTs for years and that many of my best public-ones are still publicly accessible. But I also used the sharing feature so that I could jump between my accounts, so still private access but used outside of X or Y account. For example I have one account dedicated to my business and one dedicated to personal use. Perhaps are more reasonable limit would be capping sharing at 20 people. This also means I can’t make Custom GPTs for my family anymore…

I get it. This was a huge cost for little return for OpenAI but to me it was a huge value. Something I would be willing to pay have restored. Not that I want them to start taking away features and turning them into upgrades.

10

Ditto to everything you said. I’m in a similar situation; sharing GPTs with students/clients without having to reveal all my IP like I would if I were to share a skill, or give them the instructions/knowledge documents to create a project.

11

I agree 100% I’m in exactly the same situation, I’ve been with chatGPT for 3yrs, I’ve never considered moving because of custom GPTS I don’t publish to the GPT store, I use the share private link.

When I signed up for Claude, I kept my chatGPT account (plus) because of CGPTs, I would have been gone!

And to be fair I still use chatgpt more than Claude and some of the new features are very good. But if CGPTs aren’t available to Plus then I may as well cancel my subscription, yes I won’t have the use of the other new features here, but there seems little point getting used to new features as openAi could drop them at short notice like they have with CGPTs.

Looks like I need to start backing everything up and looking for an alternative.

To everyone else not happy with CGPTs being removed. The bigger lesson here is Can you trust a company that shows little to no regard for it’s users!

12

I work with educators in the K-12 environment. Most of them are fairly overwhelmed with the responsibilities of teaching and classroom management, so introducing technology needs to be as comfortable and quick to use as possible. Custom GPTs are super easy to distribute with a link and easy to use. They function like an app; the instructions and guidelines built into the CGPT are always present. This ease of use and constant presence of the internal source-of-truth guidelines are critical for success. For parents and students, these features are even more critical.

For me, the ideal distribution flow is:

Receive link → click → start working, with all of the source-of-truth docs, instructions, etc. always present (permanence).

I’m concerned that this is disappearing with the plug-in model. Anything that requires any type of installation process is a barrier to widespread use. Any process that allows the AI to choose when to call the plug-in allows the user to escape the source-of-truth docs and instruction sets that are built in.

Because of the upcoming changes, Lesson Planning Agents, Assistive Technology Agents, and Student Study-Buddies that we have been providing free to teachers, therapists, parents, and students in a way that is easy to share and use will start having a higher response cost to try to use.

So this is my request: to keep some version of the custom GPT available, or to create a way to give the plug-in model a simple “receive link → click → start working” distribution front-end (installation is hidden), a simple “click → start working” once it has been used the first time, and in all cases a state of permanence.

13

The main issue for me is that I can no longer share a link to a custom GPT with someone. Previously i could create a GPT then share a link with a prospective client so they could check it out. Now I can not do this unless I create a public plugin however that exposes the plugin to whoever looks at it.

So I no longer have any method of creating and sharing a quck prototype with aq prospective customer.

Furthermore, I tried migrating one of my GPTs and the process failed. That is, it did not accurately reproduce the GPTs behavior. I was able to correct this with a bit of work however doing this over and over again is problematic and for complex GPTs it is a big deal. Essentially you have shut down my business. thanks a lot.

14

I am a ChatGPT Plus user who has developed substantial custom GPTs and shared work-in-progress GPTs with selected testers/users. The announced migration to Plugins appears to remove a critical capability of that workflow.

Previous workflow

Creator → privately/shared GPT → selected user can use GPT → creator retains control of implementation → creator updates GPT centrally.

This is important not merely for casual sharing, but for developing proprietary workflows, beta testing them with individual users, and iterating before broader release.

Migration result

A migrated personal plugin starts private. OpenAI’s migration documentation states that GPT sharing settings do not transfer and existing users are not automatically given access to the replacement.

I created a disposable plugin specifically to test possible workarounds.

Test 1 — Personal plugin

The plugin works correctly for its owner. Its skill and instructions are visible to the owner through the plugin-management UI.

There is no apparent personal-plugin equivalent of the previous private/user sharing controls.

Test 2 — Access-code workaround

I created an access-code gate in the plugin skill. The gate works behaviorally, but this does not solve distribution because the personal plugin itself remains private. A password therefore cannot turn a private personal plugin into something selected external users can invoke.

Test 3 — Shared conversation

I invoked the private plugin and shared the resulting conversation with another user.

The recipient could see the plugin-generated conversation and continue it. The continued conversation even retained enough context to respond to the access-code interaction.

However, when the recipient attempted to invoke functionality from the plugin, ChatGPT reported that the functionality was not available in that chat. Sharing the conversation therefore did not grant executable access to the private plugin.

Result

A workflow that was possible with Custom GPTs does not currently appear reproducible with personal Plugins:

Creator owns implementation → selected external users can use it → users cannot edit/copy implementation → creator can update it centrally.

This is potentially a severe regression for independent developers, educators, consultants, researchers, and small businesses who used Custom GPTs to distribute proprietary or work-in-progress tools without operating their own infrastructure.

OpenAI already appears to have the required access model for workspace plugins: “Only those invited.” The missing capability is an equivalent sharing option for personal plugins.

Requested solution

Add Share plugin → Only those invited to personal plugins, ideally with the same distinction Custom GPTs had between Can chat/use and Can view settings.

Alternatively, provide an unlisted plugin option in which possession of the link permits installation/use without exposing the plugin’s source files.

This would preserve the practical Custom GPT workflow while allowing the underlying implementation to move to Plugins.

Why this is urgent

OpenAI currently says Custom GPTs are scheduled to retire December 11, 2026. Existing GPTs remain usable until then, but migrated GPTs become read-only, and OpenAI itself recommends testing replacement workflows before switching.

At present I cannot safely migrate important shared GPTs because I have no equivalent way to give their existing users access to the replacements.

This isn’t primarily a request to preserve the old GPT architecture. It is a request to preserve a fundamental distribution capability during the transition to the new architecture.

15

This post expresses it totally correct. I agree 110%
I have been an ardent promoter of ChatGPT, even doing presentations to get people in my profession and in my community to start using ChatGPT. The Custom GPT feature is more than adequate and since you are not replacing the Custom GPTs entirely, this indeed looks like a targeted cutback of our existing subscription plans. Not nice, OpenAI :unamused_face:

16

what did you do to fix the behavior? I ran the same test and found the Plugin did not mirror the Custom GPT’s behavior as well. This implies the Migrate function is flawed and OpenAI didn’t do enough testing…no idea why they wouldn’t make this work better. But I think the solution for me is to build a skill (or other method) to fix the Plugin so we can (1) Press the Migrate button (2) run the skill → be done. So what type of behavior did you have to fix? I plan to work on all of this over the weekend and will probably be posting my results.

17

I had to rewite the prompts and aux files i used for the original GPT and I had chatGPT help me do this. The issue was the depth and length of the replies suddenly went from in depth analysis to just a few words. Luckily I had many previous conversations and I was able to restore the functionality by opresenting examples of the for I wanted.