Scan statuses explained

Pending, processing, completed, failed, no_brands_found — and what each means.

Last updated: 2026-05-05

TL;DR — Each query in a scan ends in one of five statuses: Pending (queued), Processing (running now), Completed (response captured and parsed), No brands found (valid response, but no brand mentioned — not an error), or Failed (model error, timeout, or unparseable response after one retry). "No brands found" is signal, not failure.

Why statuses matter

A single AI Scan can include hundreds of (prompt × model) executions. Each one has its own lifecycle. The scan-level summary tells you "117 completed, 2 no brands, 1 failed" — but to debug or interpret a specific query, you need to know exactly what its status means and what (if anything) you should do about it.

Statuses also determine billing: only certain statuses consume credits.

The five statuses

Pending

The query is in the queue, waiting for a worker. Pending rows haven't started executing yet.

What you see: a clock icon, "Queued" label.
Action: none. The queue picks it up automatically.
Credits: not yet charged.

Processing

A worker has picked up the query and is currently calling the model. Most queries spend 2–30 seconds here.

What you see: a spinner, "Running" label.
Action: none. Wait for completion.
Credits: not yet charged (charged at terminal status).

Completed

The model responded, the response was parsed, and brand/source extraction succeeded.

What you see: a green check, plus the row's data (brands, sources, sentiment).
Action: read the result.
Credits: charged at the per-query rate.

No brands found

The model produced a valid, parsed response — but no brands were mentioned at all. This is not a failure. It's a meaningful signal: for this prompt and this model, no brand made the cut.

What you see: a neutral grey badge, "No brands found".
Why it happens:
- The prompt was too generic ("what is customer feedback?")
- The model gave a definitional/conceptual answer rather than a recommendation
- The category genuinely has no dominant players the model knows about
- The model declined to recommend (rare, usually with sensitive categories)
Action: if many prompts are coming back with no brands, sharpen them. See Generate prompts.
Credits: charged at the per-query rate (the model still ran).

Failed

After one automatic retry with a doubled timeout, the query still couldn't complete or the response couldn't be parsed.

What you see: a red X, error label.
Why it happens:
- Persistent provider outage (rare)
- Content-policy block on the prompt
- Model removed or deprecated mid-scan (very rare)
- DeepSeek R1 and similar reasoning-stream models occasionally produce unparseable output
Action: check the Results page for the error message. If it's a transient outage, run a smaller scan to retry. If it's a persistent prompt issue, edit the prompt.
Credits: depends — recoverable failures are not billed; permanent failures from successful model calls are billed once (the model returned tokens). The scan summary shows the exact figure.

How statuses appear in the UI

Surface	What it shows
Scan summary card	Counts of each status (e.g. "117 completed · 2 no brands · 1 failed")
Results page	Per-row status badge with hover tooltip
Topbar progress	Aggregated progress bar while scan runs
Realtime updates	Status changes stream in live via Supabase Realtime

Worked example

Maya's 120-execution scan completes:

102 Completed — Acme appeared in 47, didn't in the rest.
12 No brands found — mostly broad "what is X" prompts where the model explained rather than recommended.
4 Failed — all on the same model, all hit a transient provider outage; auto-retry recovered most but these 4 didn't.
2 are still Processing as she opens the page; they finish 30 seconds later.

Action items she pulls from this:

Sharpen the 12 "no brands" prompts to be more recommendation-shaped ("best customer feedback tools" not "what is customer feedback").
Re-run the 4 failed ones (with the same model) to recover the data.
Use the 102 completed rows to drive this week's dashboard.

Why "No brands found" is not failure

This is the single most common confusion. The system literally invented this status because lumping it under "failed" or "completed" both produced wrong intuitions:

Lumped under "failed": users would re-run prompts that were working perfectly fine, wasting credits.
Lumped under "completed": visibility metrics would silently include zero-brand responses as if your brand "could have" appeared, distorting the denominator.

By giving it its own status, dashboards can correctly account for it: visibility = mentions / (completed responses where brands could have appeared).

Edge cases & gotchas

Status can change after you've looked: a Failed row that's actually been retried successfully will flip to Completed — refresh the page if you're not seeing real-time updates.
Mass Failed usually means a transient provider issue. Wait an hour, run a fresh scan with the same prompts.
No brands found cluster on one model can indicate that model is unusually conservative for your category — useful intel.
Permanent failures still appear in the Results table so you can see what was attempted, just with no parseable response data.

Troubleshooting

Symptom	Likely cause	Fix
Many `No brands found`	Prompts too generic	Use comparison/recommendation-shaped prompts
Many `Failed` on one model	Provider outage or model deprecation	Try a different model; retry later
`Processing` stuck >10 min	Worker batch waiting on slow model	Wait; if persistent, contact support with scan ID
Status doesn't update in UI	Realtime subscription dropped	Hard refresh
`Completed` but brand cell is empty	The model talked about your brand but the extraction didn't recognize the alias	Add an alias for that brand variant; re-run

Still need help?

Email contact@pranas.co with the scan ID and the specific row IDs that look wrong; we can inspect the queue logs.

AI search visibility metrics guide: how completed answers turn into Mention Rate, Coverage and AI Score.
My dashboard is missing data