Every listing photo goes through AI-generated and moderation checks. Cockpit shows how often those checks flag real products, which detector causes it and which threshold would cut it.
Health from €149 a month after a 14-day trial with no card.
A detector that flags too much blocks honest sellers and floods your support queue. One that flags too little lets fake or unsafe listings through. Without numbers, nobody can say which one you have.
The flag rate and confidence chart per check, drift alerts when a provider changes behaviour, and, from your trust and safety decisions, the false alarms and misses per detector.
Apply the suggested threshold in Controls, send unsure listings to a second detector, and watch the false alarms drop week by week.
Made-up data on a test account. No real customer.
With your review decisions connected, Cockpit suggests a threshold per detector that cuts false alarms with only a small rise in misses, and shows the before and after.
No. Suggestions are only applied when someone on the Control plan saves them, with a reason, and every change can be undone.
No. Cockpit works on verdicts, scores and file hashes only.