My scan failed
Most failures auto-retry. Here's what to do if a scan still fails.
Last updated: 2026-05-05
First, check the auto-retry
Most failures auto-retry once with a doubled timeout. Look at the query in the Results table — if the latest attempt is still Failed, then the auto-retry has already run.
Each failed query exposes the error reason in the row's expand panel. Common causes:
| Error | What it means | What to do |
|---|---|---|
model_unhealthy |
The provider was rate-limiting or down | Wait a few minutes; the model auto-recovers when health checks pass |
timeout |
The model exceeded the per-request timeout | Re-run with a lighter model, or wait — provider load fluctuates |
parse_error |
The model returned an unparseable response (common with DeepSeek R1) | Try a different model |
rate_limit |
You hit a per-user concurrency cap | Wait — remaining queries continue automatically |
no_credits |
Your balance reached zero | Top up or upgrade |
Step-by-step recovery
- Open the failed query to see the exact error reason.
- Try a different model — most failures are provider-specific. Re-run with a sibling model from a different provider.
- Re-run the scan from the Results page (this creates a fresh queue entry).
- Check our status page if multiple models fail simultaneously — it's likely a wider OpenRouter outage.
- If it persists across runs, contact contact@pranas.co with the scan ID (visible in the URL).
Will failed queries cost credits?
No. Failed executions are not charged. You only pay for queries that return a parsed result (or a clean "no brands found" status).
Partial scans are OK
A scan with 50 prompts × 6 models = 300 executions. If 5 fail, the dashboard still uses the 295 successes. The failed queries are flagged so you can re-run just those.