ReviewProof ⚖

How is the ReviewProof risk score calculated?

By the Ingenious AI team · Published 14 November 2026 · Reviewed 14 November 2026

Run the free review audit See pricing · See a sample provenance report

Short answer

The risk score is computed only from your imported rows: 100 × (gating × 1.0 + incentive × 1.0 + duplicate × 0.8 + AI-signal × 0.6) ÷ (2.6 × number of reviews), plus 15 if a burst is detected, capped at 100. It is an index of how much of your own file carries solicitation or fabrication patterns — a triage number for deciding where to look first, not a grade of your business and not a verdict on any single review.

What is the ReviewProof risk score?

The ReviewProof risk score is a 0–100 index summarising the density of risk signals in an imported set of reviews. The denominator (2.6 × reviews) is the maximum weighted signal load a file could carry per review, so the score reads as a proportion of the worst case. Because it is computed from the rows you supplied, you can reproduce it by hand from the report — every input is printed beside the number.

Work the formula once and it stops being mysterious

Take a file of 100 reviews with 4 gating rows, 2 incentive rows, 3 duplicate rows and 5 AI-signal rows, and no burst. The weighted load is 4 + 2 + 2.4 + 3 = 11.4; divided by 260, times 100, gives a score of 4. A small score — correctly, because 86 rows carried nothing at all. Now take 30 reviews with 10 gating and 8 incentive rows and a burst: load 18, divided by 78, times 100 is 23, plus 15 for the burst gives 38. The formula rewards nothing and punishes nothing; it measures concentration, which is why small files swing harder than big ones.

ReviewProof computes every signal from the rows you import, never from a reputation guess about your business. The risk score is 100 × (gating × 1.0 + incentive × 1.0 + duplicate × 0.8 + AI-signal × 0.6) ÷ (2.6 × number of reviews), plus 15 points when a burst is detected, capped at 100. Each flagged review in the report names the imported row that triggered the signal, so a human can open that row, read it in context, and decide. The product labels these patterns signals for human review, not verdicts — and that restraint is deliberate, because a heuristic that overclaims would be its own kind of false evidence.

Reading the score like an auditor, not a fan

Three reading rules keep the number useful. First, compare a location against its own past imports, not against other businesses — baselines differ by trade and by history. Second, always read the score with its breakdown: a 30 driven by gating says 'fix the ask', while a 30 driven by duplicates and bursts says 'investigate a supplier'. Third, remember what is not in the number: transaction backing. A low-risk file can still be mostly UNLINKED, and a higher-risk file of fully linked reviews may be the more provable profile. The score triages wording risk; the certificate answers provenance. Confusing the two is the commonest misreading.

In practice the routine takes minutes. Export or copy the reviews for the location, import them into ReviewProof by CSV or paste, and read the report top-down: total rows, flagged count, risk score, then the per-signal breakdown. Open the flagged rows themselves — the report traces each signal to its row precisely so you can do this. Decide moderation openly (kept, published, or flagged internally; there is no delete), check which reviews carry a transaction link, and publish or refresh the verified widget so the public claim matches the audited record. On Pro, export the evidence pack while the file is fresh.

Mistakes that undo the work

Quoting your risk score in marketing is a category error — it is an internal triage index, and a low one proves nothing to a customer; the certificate is the public proof. Averaging scores across locations hides the one location that needs help. And chasing a zero by pressuring reviewers' wording ('please don't say we gave you a discount') teaches customers to hide signals instead of removing the underlying practice, which is worse than the score ever was.

Two design choices keep the detection honest. First, nothing is scored until you import it: there is no scraping of your public profiles and no third-party score blended in. Second, the weights reflect what the FTC fake-review rule (16 CFR Part 465) actually prohibits — how a review was solicited weighs heaviest (gating and incentive markers at 1.0 each), while wording patterns weigh less (duplicates 0.8, AI-generation patterns 0.6). A burst of same-day reviews adds a flat 15 points because timing is visible in the dates you supply. You can reproduce every number in the report from your own file, which is exactly the point.

Worked example: one score, two very different files

A fictional worked example, written to show the numbers and the decisions in practice.

Two businesses both scored 27. The first, a physiotherapy clinic with 88 imported reviews, got there from 14 gating rows — its old intake form's wording, now replaced, with every review transaction-linked. Its action was training and a re-audit. The second, a car wash with 41 reviews, reached 27 from a burst day (15 points) plus duplicates and AI signals, with 30 of 41 rows UNLINKED. Its action was a supplier investigation and a platform dispute, evidence pack in hand. Same number, opposite meanings — which is why the report prints every component and traces every flag to its row, and why this guide exists.

Your five-action checklist

  1. Recompute the score by hand once from a report's printed components
  2. Read the per-signal breakdown before reacting to the headline number
  3. Compare each location's score against its own previous import
  4. Check the transaction-linked share alongside the score, every time
  5. Let the dominant signal decide the action: solicitation, supplier, or data quality

Frequently asked questions

What is the exact risk-score formula?

100 × (gating × 1.0 + incentive × 1.0 + duplicate × 0.8 + AI-signal × 0.6) ÷ (2.6 × total reviews), plus 15 if a burst is detected, capped at 100 and floored at 0.

Is a high score proof of fake reviews?

No. It is a concentration index of heuristic signals in your imported file. Each flagged row still needs human judgement in context.

Why do gating and incentives weigh most?

Because they describe how reviews were solicited, which is the conduct the FTC fake-review rule directly prohibits. Wording patterns are weaker evidence and weigh less.

Does transaction backing affect the score?

Not the formula — verified and unverified counts are reported separately in the summary. Read both: wording risk from the score, provability from the linked share.

Keep reading

Related guides from the ReviewProof library:

Start with the tool itself: open the free audit desk, see how ReviewProof works, compare Free, Pro and Agency plans, inspect the sample provenance report, or browse all ReviewProof guides.