Build your test set
- 8 everyday questions your users really ask.
- 1 hard case with several steps or edge conditions.
- 1 question your files should answer, and 1 that they should not (the bot should say it does not know).
Run them on the original GPT first and save the answers, before you migrate or while it is still available.
Score each answer
| Check | Pass if |
|---|---|
| Accuracy | Facts match your source files |
| Tone and format | Reads like the original; follows your format rules |
| Files | Uses the right file; does not invent answers when missing |
| Boundaries | Refuses what the original refused |
| Tools | Every connected tool works end to end |
For plugins, test two ways
OpenAI's guide suggests checking that the right skill is selected, that it follows your instructions and handles the harder case, and testing both explicit selection (@ mention, or + then More) and a normal task request. See plugin migration.
Fix and repeat
Most gaps are fixed by tightening instructions or re-uploading cleaner files. Run the full set again after each change. Only then tell users to switch; the checklist covers announcing the move.
Check your GPT before you move it
Paste your instructions, file names and Actions schema into the free checker. It runs in your browser and builds a portable bundle. Nothing is uploaded.
Open the free checkerRelated guides
Pages describing other companies' products reflect their public documentation on the date shown and can go out of date.