Scan failures

Executions can fail for reasons on your side or the provider's side. The queue shows every failure with its reason, and the queue page's stat cards show how many failed in the last hour.

Errors you can fix

1. Invalid API key: the provider rejected the stored credential. Re-save the key in API Settings.
2. No API key configured: add the credential for that platform, or deselect the platform for the topic.
3. API key out of quota or credit: the provider accounts are prepaid, so top up credit with the provider. This affects OpenAI, Anthropic, Perplexity, Gemini and DataForSEO accounts.
4. Plan quota exhausted: your monthly platform executions are used up. Add your own keys or upgrade; the quota resets monthly.

Errors that resolve themselves

1. Provider rate limit: the platform throttled requests; jobs retry automatically
2. Provider timed out or provider error: transient provider problems
3. Unreadable provider response: the answer arrived malformed; a retry usually fixes it

Not a failure

1. AI Overview backoff: prompts where Google shows no AI Overview are checked weekly instead of every run. These executions are skipped on purpose and marked as such.
2. Empty answers: an answer with no text is a valid result, not an error.

Staying informed

The scan-failure email notification tells you when failures have a cause you can fix. Enable it in the notification preferences; silent failures are the main way tracking gaps go unnoticed.