A Custom GPT isn’t a standalone service — it’s a configuration layer on top of your ChatGPT account: system instructions + knowledge files + optional actions (external API calls). Any one of these layers can fail in a way that looks like “the GPT is broken” — but the symptoms are similar enough that correctly identifying which layer is the key to fixing it.
This article is organized by symptom → cause → fix.
Symptom map
| Symptom | Most likely cause |
|---|---|
| GPT card opens blank / spinner forever | Browser cache holds an old manifest / network block |
| Normal chat works but action returns “request failed” | Action backend is down / auth expired |
| ”This GPT is no longer available” | Creator deleted it, or OpenAI removed it |
| Only works for the creator | Set to Private or restricted access |
| Switching Web → App drops GPT context | Some GPTs have App-side compatibility issues |
| ”This GPT requires Plus” | Subscription expired / on Free tier |
Common causes
In rough order of frequency:
1. Browser cached an old GPT manifest
The GPT’s instructions + action schema get cached by the browser. After the creator updates them, you may still see the old version — out of sync with the backend — and the GPT goes blank or behaves strangely.
How to verify: hard refresh (Cmd/Ctrl + Shift + R). If it works after, it was the cache.
2. Current account doesn’t have access
If the creator set the GPT to Private or “Anyone with the link” without adding your account to the allowlist, you’ll see “GPT not found” or get bounced.
How to verify: have the creator send you the link — if you still can’t open it, your account doesn’t have access.
3. Action’s backend API is down / auth failed
A Custom GPT’s actions are HTTPS endpoints the creator hosts elsewhere. If their service is offline, the SSL cert expired, or the API key was revoked, the action returns “request failed” — but the GPT itself (the no-action parts) still works.
How to verify: in the GPT detail page bottom, click “View source / Actions” to see the manifest URL. Open that URL in a browser and check whether it returns 200.
4. Creator deleted / OpenAI removed
Creator deleted → “GPT is no longer available.” OpenAI removed for policy violation (impersonation, deception, copyright) → same message.
How to verify: the error explicitly says “no longer available” or “removed.”
5. Requires Plus, subscription lapsed
Some GPTs internally use GPT-5 / image / browsing — Plus / Team only. Free accounts trying to open them get an upgrade prompt.
How to verify: the error mentions “requires ChatGPT Plus” or “upgrade to continue.”
6. Browser extension / network blocking
Ad blockers may strip the GPT iframe; corporate firewalls may block the action’s external API.
How to verify: incognito + 4G hotspot. If that works, extensions or office network are the issue.
Shortest path to fix
30-second checks first:
Step 1: Hard refresh + incognito
- On the GPT page, press Cmd/Ctrl + Shift + R (bypass cache)
- Open an incognito window (extensions disabled), sign in with the same account, open the GPT
- Still blank → jump to Step 3
If incognito works → cache or extension issue:
- Cache: DevTools → Application → Storage → Clear site data
- Extensions: add chatgpt.com to the whitelist in uBlock / AdGuard / etc.
Step 2: Try the app / a different device
If Web fails, try the mobile ChatGPT app with the same account. App works = desktop browser layer issue (back to Step 1). App also fails → Step 3.
Step 3: Confirm account permissions + tier
- Read the error for “requires Plus” — if so, upgrade
- Read for “GPT not found” — may be a permissions issue, ask the creator for a fresh link
- Avatar → Settings → Subscription to confirm tier status
Step 4: Diagnose action failures
If the main chat works but the action returns “request failed”:
- Ask the GPT “show me the action URL you are calling” — the model will tell you
- Copy the URL to Postman / a browser tab and verify the backend returns 200
- Backend down / 401 unauthorized → only the creator can fix it
If you are the creator:
- Verify the manifest URL is reachable
- Check API key / OAuth config hasn’t expired
- Action drawer → “Test” button to dry-run
Step 5: Replace it or build your own
If the GPT is reliably broken and you depend on it:
- Search the GPT Store for an alternative (entries marked “official” are more stable)
- Copy the GPT’s instructions and paste them as a system prompt in a native chat
- Build your own Custom GPT internalizing the core prompt
Prevention
- Pin only GPTs you use long-term from trusted creators — third-party GPTs can disappear at any time
- Save the prompt of critical-workflow GPTs outside the GPT (Notion / Apple Notes) so you can rebuild
- For your own GPTs, version the manifest URL (
/v1/manifest.json) and bump the version when changing the schema instead of overwriting - Periodically Export your important GPTs’ instructions — OpenAI occasionally corrupts GPT data
- For business-critical actions, don’t rely solely on a GPT call — have independent monitoring on the backend
Related
- ChatGPT projects guide
- ChatGPT slow response
- ChatGPT beginner guide
- ChatGPT prompt improvement
- ChatGPT model selection guide
Tags: #ChatGPT #Debug #Troubleshooting