I have not evaluated other platforms for this, and I would rather not have to: my clients already pay OpenAI for Plus to use my GPTs, several with over 100 conversations here, and moving them elsewhere means every one of them changing platform, account and history. The cheapest path for everyone, OpenAI included, is an unlisted plugin link for people outside the workspace. That is why I keep asking for it in writing, and why I would suggest everyone here does the same: one support ticket, one paragraph, the same request, quoting this topic number. If someone does test alternatives, by all means post what you find, but the ask in this thread should stay clear and simple.
Update: OpenAI support answered my three questions in writing this morning, itemized, not a template. (1) By current documentation, no: plugins have no option equivalent to “Anyone with the link” for people outside the workspace; the documented sharing options are workspace-only (invited members, workspace link, workspace directory). (2) The documentation describes no alternative “without the public directory” to keep access for users outside the workspace after Dec 11; migration does not carry the GPT’s permissions, the migrated plugin starts private, and broad access requires making it public through the submission process. (3) They cannot answer “yes” in general on Free/Plus: the directory exists on all plans, but installing and using a specific plugin depends on plan, account, region and plugin capabilities, with no documented guarantee for Free and Plus in every case.
So the position is now clear and in writing, and it matches what the UK creator above was told. I replied asking for four things: (a) the unlisted install link for migrated plugins, usable by people outside the workspace on their own Free/Plus accounts, logged as product feedback and quoting this topic; (b) a priority or partner review lane for builders migrating GPTs with external customers; (c) until (a) exists, written confirmation for my workspace that Free and Plus accounts can install and use my plugin once approved, and that link sharing on Business stays until Dec 11; (d) a named contact with authority. I will post whatever comes back. If you sell access to GPTs by link, the most useful thing you can do this week is send support the same one-paragraph request, quoting topic 1399984.
I completely agree with you. I haven’t looked at anything else either and don’t have the time to do so; we’ve all learned how to build some incredible things here using Custom GPTs which have benefited our businesses and those of our clients, so I am hopeful that OpenAI will do as you suggested, just create the ability to create an unlisted share link.
I´m devasted with this new update regarding Custom GPTs migration to plugins ! I´ve being creating and using them for my business for years! Some of them were created for my customers. Some of them I got from third parties and can´t contact the creators to ask for the plugin versions. I understand OpenAI´s need to improve their business but there are so many issues regarding this change that make me think about how trustful is to sign in paid tools with this company ! I prefer to keep my main LLM here but if it´s becoming unreliable, I´ll need to revise my business strategy ! Hope OpenAI revise this decision and/or the way that it´s being implemented.
I bet, Open AI will give a sh*t answering us all that questions and requests. Our requests are based on logical, well-reasoned, and very sound arguments, and yet I fear that OpenAI won’t care at all. The days when OpenAI listened to its community seem to be over once and for all -just look at the shutdown of GPT-4. There are simply models and concepts that are really damn good just the way they were - why not just keep them? Even Midjourney still offers the option to use ALL models since V1.
Besides, I’m probably not the only one who feels like this whole thing with all the different models and “fast/medium/supermegaboomhigh” is getting completely out of hand. There are TOO many tiers, TOO many options, TOO many features. I still don’t understand what the advantage of “skills” and “plugins” is over a simple, stable GPT. It just doesn’t make any sense. I suspect it’s simply driven by cold, hard financial interests. OpenAI probably doesn’t care at all about the “small” users any more.
It´s just extremely frustrating and after many years using ChatGPT I´m really thinking of leaving this mess and maybe quit using ai at all.
YES! Same here. The amount of time and work that’s going to be lost is massive.
You are right, everybody are right, a
And now what?
I think what will happen is that open ai will lose lots of subscribers, as businesspeople will migrate their subsctiptions to other platforms, migrating to private plugin, and submitting your plugin for review if you can get past the technical hurdles is like asking every owner of a established Restaurant to close their doors, change their business model, stay shut down indefinitely until they are are approved, it is a very snobbish way of treating the same people that helped GPT develop its expertise, there is no doubt that some of the ideas created by custom gpt’s will be incorporated into their own and they learned a great deal from developers and now treat them as second class citizens not worthy of continuing their business, a very shameful act on the part of management that made this decision. I hope they come down from the heavens and face the reality of very poor decision making.
no idea! Feels like there is nothing we can do. OpenAI just does what it wants.
Wow, this is brilliant, thank you so much for sharing your workaround. So you’re sharing access to the PROJECT, that’s very interesting. I’m going to have to experiment with your idea, too.
It’s just ridiculous how many hours we all spent creating these things.
Only issue I realised quickly is that all members of the project see conversations. So if you want segregation you have to create multiple projects (one per user), or if you can do collaboration/ok to see past then shared is fine. Def not perfect, but also realizing there way Plugins work to bring the skills and tools together is pretty decent. You could then gate it all with OAuth on your tools.
IE: Skills indicate you have to call “initialize”. And initialize tool establishes OAuth so you can control user access more formally.
Why are we being forced to migrate our custom GPTs to plugins? Can anyone explain the reasoning behind this?
My custom GPTs used to work perfectly, but the converted versions do not. For instance, I spent over 10 incredibly stressful days trying to get a single plugin to work properly—and that was based on a simple custom GPT with under 3,000 lines of instructions and no knowledge files. How are we supposed to convert even more complex GPTs?
Previously, if I needed a quick one-line fix, I would just open the editor and change it. Now, I have to talk to a chatbot, convince it, and practically fight with it just to edit a single line—and then cross my fingers that it actually only changes that one line (since it rarely follows instructions properly).
Furthermore, plugins are extremely slow, and they behave differently in chat mode versus work mode.
I could accept all of this as just requiring more time and effort. However, my biggest concern is privacy: why are all the plugin instructions and commands suddenly made public in the chat?
This means years of hard work are completely exposed. If I put dedicated effort into developing a solution, am I now expected to just hand my intellectual property away for free inside a plugin?
It’s one step forward, two steps back. I successfully published one converted plugin, but now the option to publish another has vanished (the “skills only” plugin option is completely gone).
I know this sounds more like a rant than a technical question, but please realize that the current reality of this system does not match the official documentation or the original proposal at all.
I used your own Chatbot to help me write this but why did you decide to Change Custom GPTs to “Plugins” - and confuse the hell out of everyone? Already calling your version of “CoWork” to “Work” was bad enough. Make GPTs work like Plugins - add the features, but don’t throw away their custom icons and the ability to distinguish a converted GPT to a Plugin. I don’t know give all the old custom GPTs a new name separate from the ones developed after Dec 11th. GPTPlugs & then your new Plugins instead of mashing everything together and making a mess for advanced users. So here is your own Chatbot solving your mistake: Enjoy!
The migration from Custom GPTs to Plugins, combined with the introduction of Projects, has removed critical metadata that many advanced users rely on to manage long-term knowledge bases.
A ChatGPT conversation is not simply a conversation. For many creators, researchers, developers, artists, and prompt engineers, it is a permanent project artifact.
Today, there is often no way to determine what created a conversation after it has been moved into a Project.
This represents a significant workflow regression.
The Core Problem: Loss of Provenance
The missing feature is not simply a missing icon.
The missing feature is conversation provenance.
When reviewing a conversation months later, users need to know:
- Which assistant created this conversation?
- Which Custom GPT or Plugin was used?
- Which ChatGPT model answered the questions?
- Which capabilities were enabled?
- Which knowledge base was being used?
Without this information, historical conversations lose a significant amount of context. For example: Real Workflow Example
Suppose six months from now I search for:
“Transparent lace preservation”
I may find several conversations.
However, those conversations could have been created using completely different assistants:
- Spirit’s Super Nano Prompter
- Spirit’s Visual Prompt Copilot Agent
- Creature Prompt Assistant
- Photoshop Assistant
- Standard ChatGPT
Those assistants contain different instructions, different knowledge files, different prompting methodologies, and different reasoning approaches.
Knowing which assistant produced the solution is essential context.
Without that information, the conversation becomes much less useful as a long-term reference.
Regression #1: Projects Hide Assistant Identity
Moving conversations into Projects improves organization.
However, Projects also remove visible assistant identity.
Projects tell users where a conversation lives.
They no longer tell users what created it. Those are different pieces of information. Both are important.
Regression #2: GPT Identity Appears Lost
Previously:
- Custom GPT avatar
- GPT identity
- GPT-specific visual recognition
were visible throughout the workflow.
After migration:
- custom GPT avatars are replaced with generic cube icons
- GPT identity is no longer displayed
- historical conversations no longer identify their originating assistant
The visual cues that allowed rapid recognition have disappeared.
Regression #3: Model Information Has Also Been Lost
Historical conversations should preserve the model used to generate responses.
For Example: Created with
Spirit’s Super Nano Prompter
Model:
GPT-5 Thinking
Date:
April 4, 2026
_________________________________________________
Months later, users often need to know:
- Was this GPT-5?
- GPT-5 Thinking?
- GPT Image?
- Standard ChatGPT?
- A Custom GPT?
- A Plugin?
Currently this information is not easily discoverable.
Regression #4: Migration Removed Important Historical Context
The migration appears focused on preserving:
- Instructions
- Knowledge Files
- Conversation Starters
However, the workflow history appears to have lost:
- originating GPT identity
- originating Plugin identity
- assistant avatar
- visible model information
- conversation provenance
For long-term knowledge management this historical metadata is just as valuable as the conversation itself.
Regression #5: Third-Party Extensions Lost Metadata
Extensions such as Superpower ChatGPT previously surfaced valuable metadata including GPT avatars and assistant identification.
After migration:
- cached GPT avatars disappear
- broken-image placeholders disappear after Project moves
- assistant identity can no longer be recovered visually
If the underlying metadata has been removed or is no longer exposed, extensions cannot restore this functionality.
If the metadata still exists internally, exposing it again would allow extensions to continue supporting advanced workflows.
Why This Matters
This is not simply about aesthetics.
It affects:
- researchers
- developers
- educators
- writers
- AI artists
- prompt engineers
- technical documentation
- long-term knowledge management
Many users maintain hundreds or even thousands of conversations spanning years.
Those conversations function as a searchable knowledge library.
Removing authorship and creation metadata reduces their long-term value.
It is analogous to removing:
- the author from a research paper
- the camera information from a photograph
- the software version from a Photoshop file
- the compiler version from source code
The content still exists, but critical context has been lost.
Suggested Improvements
1. Add a “Created With” section
Every conversation should contain immutable metadata such as:
Created With
Assistant:
Spirit’s Super Nano Prompter
Type:
Plugin
(formerly Custom GPT)
Model:
GPT-5
Knowledge Files:
Enabled
Date Created:
April 4, 2026
- Preserve Assistant Avatars
If a custom GPT had a unique icon, preserve it historically.
Historical conversations should retain the icon that existed when the conversation was created.
3. Preserve Model Information
Every conversation should permanently record:
- model name
- model version
- reasoning mode (if applicable)
- image model (if applicable)
This metadata is essential for reproducibility and historical reference.
4. Preserve GPT-to-Plugin Lineage
After migration, maintain a permanent record such as:
Originally Created With
Spirit's Super Nano Prompter
(Custom GPT)
Migrated to Plugin
September 2026
This preserves continuity and avoids confusion during platform transitions.
5. Expose Metadata to Extensions
If assistant identity still exists internally, expose it through supported APIs or metadata so trusted browser extensions such as Superpower ChatGPT can continue providing advanced organization and retrieval features.
6. Generate a Migration Report
Each migration should produce a report summarizing what was preserved and what changed.
Example:
Migration Complete
✓ Instructions
✓ Knowledge Files
✓ Conversation Starters
✓ Permissions
✓ Historical Conversations Preserved
Former GPT ID:
...
New Plugin ID:
...
Public URL:
...
Migration Date:
...
This would eliminate uncertainty and provide users with a permanent migration record.
Sou do Brasil também. Alguma novidade/solução até o momento?
News.
The documentation and plugin page changed: Upload and submit your plugin – Plugins | OpenAI Developers
My friend, this what MCP is for 
I think most of the rants have been the fact that Plugins do not equate to CustomGPTs, not apples to apples.
yes and you must understand for non teckies, they have no idea how to use MCP, and on top of that the plugin creator and lack of “proper” menu type editing system (more or less (prompt type cli driven now)), doesnt equate to what we had before.
once again i will point out the legalaties of selling service “b” and then downgrading people to a lower service without proper notice and for a lot of us mid “contract”, is totally against eu service laws!
its rich they are trying to “protect us” by their guidelines, but not protect us from their completely illegal actions.. anyway no matter theirs a 10 % chance they are going to kill us all anyway eh sam? my auant nelly you are trying to project us! “abundance of caution”, more like abundance of Enterprise gaslighting" – Telling individual users they are being “protected” while actively stripping away the value they paid for to cater to big corporate contracts.
heres the message ive recieved from open ai regarding my saying to people dont dispare even though open ai have done everything possible to scupper probably thousands if not tens of thousands of small developers, i have found a way to convert most custom gpt’s to plugins…
I said this in a previous post and gave clues how to contact me, as i think direct comunication may not be allowed in this forum.
so more proof that open ai dosnt care a hoot about any of us!!! or the pain they have caused by their completely illegal actions.
"Enterprise gaslighting" – Telling individual users they are being “protected” while actively stripping away the value they paid for to cater to big corporate contracts.
Hello,
This is an automated message from OpenAI Developer Community to let you know that your post was hidden.
Your post was flagged as spam: the community feels it is an advertisement, something that is overly promotional in nature instead of being useful or relevant to the topic as expected.
This post was hidden due to flags from the community, so please consider how you might revise your post to reflect their feedback. You can edit your post after 30 minutes, and it will be automatically unhidden.
However, if the post is hidden by the community a second time, it will remain hidden until handled by staff.
For additional guidance, please refer to our community guidelines.