Short answerIf a bot using your own API key stops answering, the usual causes are a wrong or revoked key, no credit or billing set up with the provider, or hitting a rate limit. Check the key and billing in your provider account.
Common causes
| Symptom | Likely cause | Fix |
|---|---|---|
| Invalid or unauthorized | Key mistyped, revoked or for another provider | Create a new key and paste it again; check the provider setting |
| Quota or billing error | No credit or payment method with the provider | Add billing in your provider account |
| Rate limit | Too many requests too fast | Wait, or raise limits with the provider |
| Model not found | Model name typed wrongly or unavailable to your account | Clear the model box to use the default, or use a model your account can access |
With GPT Migrator hosting
Open the bot in your account, pick the right provider (OpenAI or Anthropic), paste the key, and leave the model blank to use the default. The preview of a few replies works without a key; after that the bot needs yours. See how hosting works.
Never share your key publicly; see protecting a public bot.
Need your GPT running again without ChatGPT?
GPT Migrator can host your instructions and files as a standalone chat page for $47 per month. You bring your own OpenAI or Anthropic API key. It answers from your instructions and files and does not run GPT Actions.
See how hosting works