Hey Reader!
AI tools are going to keep changing.
Names will change. Features will disappear. New formats will replace old ones. And sometimes the platform will make a decision before you were planning to make one yourself.
That’s exactly why the most valuable part of your custom GPT deserves to live somewhere else too.
The real asset is the method you built into it: your instructions, decisions, examples, reference material, standards, and the way you expect the finished work to look.
This week, we’re protecting that.
IDEA SPARK
A useful AI workflow should make sense even when the tool running it is temporarily out of the picture.
- Can you explain what your GPT does without opening it? You should be able to describe the problem it solves, who it helps, what information it needs, and what it produces.
- Do you have copies of the important pieces? That includes your instructions, knowledge files, examples, decision rules, output requirements, and anything else the workflow depends on.
- Could you test a replacement? Keep several familiar prompts and examples so you can compare the old workflow with whatever replaces it.
If any of those answers feel a little shaky, you’ve found a workflow worth documenting before you need it.
NEW ON BLOG
My newest article digs deeper into what creators should save before migrating a custom GPT.
- Understand what actually needs protecting. The workflow inside your GPT may be more valuable than the GPT itself.
- See what migration may miss. Instructions, reference files, connected actions, behavior, and output quality all deserve a closer look.
- Use test prompts to compare versions. Familiar prompts can help you spot subtle differences after a workflow moves.
Read Your Custom GPT Is Retiring: What to Save Before You Migrate on Medium.
The goal is simple: make sure the useful thinking you built doesn’t disappear just because its container changes.
RESOURCE SPOTLIGHT
I created the Custom GPT Migration Checklist to help you figure out what needs attention without turning migration into another 47-tab research project.
- Track A is for GPTs you created. You’ll identify what each GPT does, save the important pieces, document dependencies, and prepare it for migration.
- Track B is for GPTs you use but don’t own. You’ll document how you currently use the tool so you can protect the workflow even when you can’t access its underlying instructions.
- Both tracks help you prepare for what comes next. You’ll know what you rely on, what needs to be backed up, and what should be tested after migration.
You can download the free Custom GPT Migration Checklist here and start with the GPT you would hate to discover missing tomorrow morning.
PROMPT DROP
Use this prompt with a custom GPT you created to pull the method out of the tool and document it in a format you can actually reuse.
- Start by describing the GPT in detail. Include its purpose, audience, instructions, knowledge files, actions, and the kinds of outputs you normally expect.
- Ask AI to map the full workflow. You want the inputs, outputs, decision points, standards, required files, integrations, and representative test prompts documented.
- Require uncertainty to be flagged. The model should identify anything missing instead of quietly filling gaps with assumptions.
Copy and paste this prompt:
I created and own the following custom GPT:
[Describe its purpose, audience, instructions, knowledge files, actions, and typical outputs.]
Help me document it as a portable AI workflow that can be rebuilt in another approved format.
Create:
- A one-paragraph workflow summary.
- The required inputs and expected outputs.
- The decisions and quality standards built into the process.
- A list of required reference files and integrations.
- Five representative test prompts with pass/fail criteria.
- A migration checklist separating automatic, manual, and uncertain items.
- A list of information that should be backed up before migration.
Flag anything I have not supplied. Do not invent missing instructions, claim ownership of third-party material, or assume connected actions will transfer automatically.
Once you have this documented, you’ve created something much more useful than a collection of copied instructions. You’ve created a rebuildable business process.
Before you go, here’s the question I want you to think about:
Which custom GPT would be hardest for you to replace tomorrow?
Start there.
Document the method. Save the pieces. Give your best workflows somewhere safe to live when the next platform change comes knocking.
Small steps create big shifts. 👑
Johnna