Custom GPTs Retire Dec 11: Small Business Migration Guide
If your small business built workflows on Custom GPTs — a support bot trained on your docs, a brand-voice writer, a lead qualifier — mark December 11, 2026 on your calendar. That’s the day OpenAI says Custom GPTs “stop running.”
OpenAI announced the retirement in its September 11 release notes and confirmed the timeline in its official migration FAQ. The replacement is Plugins: reusable instruction sets that can bundle reference materials, connected apps, and tools into a single workflow. Whether that’s a clean upgrade or a forced rebuild depends on how much you invested in GPTs — and how quickly you act.
The Timeline That Matters
For Enterprise workspaces, the milestones are already locked in. Other ChatGPT plans are expected to follow the same schedule, though OpenAI says all dates are subject to change.
- September 11, 2026 — Admin notice sent to Enterprise organizations.
- September 17, 2026 — Target date for the in-product migration experience and user banner. Note: a target, not a guarantee for every workspace.
- September 25, 2026 — Creation of new Custom GPTs ends for affected workspaces. Publish any drafts you need to migrate before this cutoff.
- December 11, 2026 — Scheduled retirement. Custom GPTs stop running entirely.
Personal accounts (Free, Go, Plus, Pro) can no longer create or publish new GPTs at all — they can only keep using existing ones until retirement. Creation is now limited to Business, Enterprise, and Edu workspaces.
What Actually Happens to Your GPTs
The migration path moves your GPT’s instructions, knowledge files, and configuration into a Plugin. But there are catches small businesses should know about:
You can only migrate GPTs you own. Permission to use someone else’s GPT doesn’t give you permission to migrate it. If your team relies on a third-party GPT — from a vendor, a consultant, or a public listing — you need to ask the builder whether they plan to migrate it. If they don’t, your workflow breaks on December 11.
Your admin controls the gate. For Business and Enterprise workspaces, migration requires that plugins are enabled and the GPT is published. If you don’t see the migration option, that’s the first thing to check.
Back up before you touch anything. Download your knowledge files and copy your instructions before starting migration. Once a GPT stops running, retrieving that material gets harder.
Why OpenAI Is Doing This
Plugins are part of OpenAI’s broader push toward agentic workflows — the same direction we covered when OpenAI launched its ChatGPT for Small Business program and again with GPT-5.6 and ChatGPT Work. A Custom GPT is essentially a wrapper: instructions plus files plus optional actions. Plugins do the same job but slot into ChatGPT Work and Codex surfaces where agents can chain steps, call tools, and use connected apps — closer to how businesses actually want AI to operate.
For simple use cases, migration should be close to one-click. For GPTs with custom actions or complex knowledge bases, expect some rebuilding.
What Small Businesses Should Do Now
- Inventory your GPTs today. List every GPT your team uses, who built it, and what business process depends on it.
- Identify third-party dependencies. Email the builders of any GPT you rely on but don’t own. Ask for their migration plan.
- Back up everything. Export knowledge files and instruction text before touching the migration tool.
- Migrate early, don’t wait for December. The migration experience started rolling out September 17. Early migration gives you time to test and fix breakage while your GPTs still run.
- Reassess the platform fit. If your “GPT” is really a business process — intake, triage, follow-up — a chat-wrapper was always a compromise. Purpose-built AI agents are more durable than any single vendor feature.
That last point is the real lesson here. Platform features come and go — Custom GPTs lasted barely three years. The workflows you built on top of them are the actual asset. We’ve been through this migration cycle with clients enough times to know: the businesses that treat AI as a stack of vendor features rebuild every 18 months. The ones that treat it as infrastructure don’t.
The Bottom Line
December 11 is the hard deadline, but September 25 is the one that will bite people — after that, no new GPTs, which means no clean migration for anything still in draft. If Custom GPTs are part of your daily operations, spend an hour this week on the inventory and backup steps. Future-you will be grateful.
Want a migration plan built for your actual workflows — not just a chat bot swap? SQUIDBOT builds and runs AI agents that survive platform changes. Book a free build session and we’ll map your GPT dependencies in one call.
