Trust & moderation
Every new review gets a risk score between 0 and 1 from a rules engine that runs instantly, followed by an automated text check. What happens next depends on the score.
| below 0.3 | Published. |
| 0.3 – 0.6 | Published with a “verification pending” tag. We ask the reviewer for proof of purchase. |
| 0.6 – 0.8 | Held for a human moderator. Not public until cleared. |
| 0.8 and above | Not published. Reason code auto_risk, one appeal to a human. |
Signals we look at
- Brand-new accounts, bursts of reviews from one network address, and near-duplicate text.
- Reviewers whose email is on the business's own domain (counted as insiders, excluded from the score).
- Extreme ratings with almost no text.
- The reviewer's history on the platform, which lowers risk for long-standing accounts.
- Reviews written from an invitation link, which lowers risk and marks the customer verified.
What moderators can and cannot do
Moderators work from a separate admin system. Business accounts cannot remove or hide reviews; they can reply publicly and flag a review for a moderator with a stated reason. Sales staff have no moderation access at all.