Thank you, Darin — that is exactly what I was hoping existed. I have not seen that migration banner on my own GPTs page yet, so it appears the rollout is not uniform across accounts/workspaces. This is encouraging. I’m especially interested in what the migration preserves: instructions, Knowledge files, capabilities, actions, and behavioral fidelity. We have been independently testing exactly those questions with a preservation workflow built around Capture → Preserve → Verify. Your reverse-engineering experiment on someone else’s GPT is fascinating too. I’d be very interested to hear what survives once you have a working plugin version.

@Jason West 👍I have a Plus account, and the option showed up last night. All 21 GPTs are now plugins with no errors found during migration. Plugin testing to follow. I do wish OpenAI had been more transparent about their intentions; it would have saved me a lot of frustration. 😃

@Damien Rothstein Damien, yes — that is much closer to the problem I am actually trying to solve. I am becoming less interested in “How do I copy this GPT into Platform B?” and much more interested in: How do I separate the intellectual property from the container? We recently tested this with one of my GPTs. Instead of preserving only the prompt, we created a reconstruction packet containing the Instructions, Knowledge, operating boundaries, security rules, functional QA tests, prompt-injection tests, and a reconstruction manifest describing what must still work after a rebuild. That exercise changed how I think about migration. The real asset is not the GPT shell. It is the combination of method + knowledge + workflow + guardrails + expected behavior + tests. Your Zapier/Make → n8n analogy is a good one. Moving into another proprietary chatbot builder may solve today's distribution problem while simply creating tomorrow's lock-in problem under a different logo. So the architecture I am increasingly interested in is: Extract once → preserve in a platform-independent form → deploy to multiple containers → regression-test each implementation. If that is the infrastructure you are building, I would be very interested in comparing notes as it develops.

Damien, yes — I’d be happy to help. What you are describing is very close to the architecture I’ve been hoping someone would build: the intellectual property lives independently of the deployment container. I can probably be most useful on the migration/validation side rather than the infrastructure side. I currently have 21 focused GPTs, and we recently took one of them through a full preservation exercise: Instructions, Knowledge, protocols, security rules, functional test cases, prompt-injection tests, and a reconstruction manifest. That gives me a fairly concrete test case for questions like: - What needs to be extracted from the original GPT? - What must remain behaviorally identical after migration? - What can be normalized into a portable format? - What should remain platform-specific? - How do we know a redeployed tool is actually equivalent rather than merely similar? Point me toward the data/help request and I’ll take a look. 😊 I also run a LinkedIn group called Custom GPTs to Apps focused specifically on this transition from platform-bound GPTs to portable tools and apps. Your work sounds very much aligned with that discussion, so you’d be very welcome there as well.

I like the phrase “decision pack, not a verdict.” That matches how I try to evaluate tools now. I also find it useful to separate findings into three buckets: FOUND — verified from a reliable source INFERRED — reasonable conclusion, but not directly confirmed MISSING / UNKNOWN — something important that still needs an answer I would add one more section to the pack: What would make us walk away? That forces the evaluation to identify deal-breakers before enthusiasm takes over. So the sequence becomes: Need → requirements → evidence → gaps → supplier questions → recommendation → walk-away conditions. It makes a purchase decision much less dependent on whichever demo looked best that morning.

@Leanne O'Connell I’m the exception—I’ve spent decades collecting software and own thousands of dollars’ worth of shelfware. :-) Which may be why I now think attention is the more expensive currency. The purchase price is visible. The real cost shows up later in setup, learning, maintenance, switching, and the mental overhead of remembering why we bought the thing in the first place.

@Jason West Case Number: 13587644 Hi, Thank you for reaching out to OpenAI Support. Thank you for clarifying. You’ve asked several specific questions, so I want to answer each one directly. Yes — Custom GPT creation is still supported in ChatGPT Business, Enterprise, and Edu workspaces. The currently documented restriction applies to personal Free, Go, Plus, and Pro accounts. For your 21 existing GPTs on personal Plus, your current Share dialog shows “Only me” and no longer offers “Anyone with the link.” Based on the behavior you have documented, we cannot confirm that link sharing remains available for those existing personal-account GPTs. Regarding ChatGPT Business: Business workspaces support Custom GPT creation and sharing capabilities, subject to the workspace’s settings and permissions. However, we should not promise that moving your existing GPTs to Business would automatically restore “Anyone with the link.” That specific sharing behavior depends on the destination workspace and should not be treated as guaranteed. Separately, you have demonstrated that your personal Plus account still allows you to create and use a new GPT, including g-6a89f2f644ac819188832f41efc7809c, even though the documented personal-account behavior says new creation is unavailable. Because that behavior differs from the documented expectation, we cannot promise that the Create capability currently visible on your Plus account will remain available. For planning purposes, it would be safest not to rely on that undocumented capability continuing. In short: - Personal Plus: existing GPTs may remain available, but we cannot confirm “Anyone with the link” sharing from the behavior currently shown on your account. - Business / Enterprise / Edu: Custom GPT creation remains supported, with sharing controlled by workspace settings and permissions. - Your current Plus Create button: it does not match the documented expected behavior, so we cannot guarantee it will remain available.

Case Number: 13587644 Hello, Thank you for contacting OpenAI Support. Thank you for following up. We understand that you’re trying to determine whether ChatGPT Business provides a workable way to continue distributing your GPTs by direct link, including whether “Anyone with the link” can be enabled, who can use those links, and whether your 21 existing GPTs can move to Business without being rebuilt. For your first question, ChatGPT Business can support “Anyone with the link” when public-link sharing is available under the workspace’s settings and permissions. As a workspace Owner, you can manage GPT settings, but we would not want to guarantee that this sharing option will appear in every circumstance, since the available sharing levels depend on the workspace configuration. For your second question, when “Anyone with the link” is available and selected, access is not limited to members of your Business workspace. Any eligible ChatGPT user with the link can access the GPT after signing in. A separate paid subscription is not required solely to use the GPT; Free users can also use GPTs, subject to their plan’s applicable usage limits. For your 21 existing GPTs, you would not necessarily need to rebuild them individually. ChatGPT supports merging a Personal workspace into a Business workspace, and GPTs from the Personal workspace are included in that migration. Before choosing this option, please note: - The Personal-to-Business merge is permanent. After migration, the Personal workspace is removed and the migrated data becomes part of the Business workspace. - The documentation confirms that the GPTs themselves migrate, but it does not guarantee that their existing Personal-workspace sharing links or sharing settings will remain unchanged. For that reason, after migration we recommend reviewing each migrated GPT’s Share settings in the Business workspace before distributing its link again. You can find additional information in the Help Center articles Sharing and publishing GPTs, GPTs in ChatGPT, and Managing workspace lifecycle and migration in ChatGPT Business.

This is sound advice, although my own experience developed somewhat differently. Rather than beginning with a finished set of Custom Instructions, Quill (my AI partner) and I discovered our working preferences through actual projects. When something proved consistently useful, we decided where it belonged. Broad preferences that should apply everywhere belong at the global level. Personal details and recurring preferences can be remembered. Project-specific context remains with the relevant Project. We also use two additional layers that may be less familiar: • A canon records what must remain true. This might include a project’s purpose, voice, values, recurring themes, terminology, or boundaries that should not be violated. • A protocol records how the work should be done. This might include the order of steps, review standards, formatting rules, or a process that has repeatedly produced good results. For example, a canon might preserve the identity and behavior of a recurring narrative voice. A protocol might require drafting first, polishing second, and checking continuity before finalizing the piece. That distinction has helped us avoid turning Custom Instructions into one enormous rulebook. I would describe our approach this way: • Custom Instructions establish the standing relationship. • Memory preserves useful continuity. • Projects hold the context of a particular body of work. • Canons preserve what must remain true. • Protocols preserve methods that have repeatedly worked. Most of ours were not invented in advance. They emerged from the work, were tested, and were formalized only after they proved valuable. Your post also suggests one useful practice I may adopt: periodically reviewing the global instructions to make sure they remain broad, stable, and free of material that belongs somewhere else.