ReviewProof ⚖

How do you use the reviewer-fraud graph in practice?

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

After each import on Pro or Agency, open the graph view and work three lists: fingerprints flagged at other businesses that appear in your file, your own repeat fingerprints with flag-worthy patterns (signal clusters, tight date groupings), and fingerprints present at many businesses without flags (usually just active genuine reviewers — context, not suspicion). Judge each flagged-elsewhere row on the full combination in your own file — transaction link, wording signals, timing — flag internally with the combination logged, and never attempt to identify a fingerprint or another merchant.

What is the reviewer-fraud graph?

The reviewer-fraud graph is ReviewProof's pooled analysis layer: per-merchant, it relates reviewer fingerprints (salted hashes of normalised handle plus source) to the reviews they wrote for you; across merchants, it pools only the fingerprint, counts and flag state, so a fingerprint flagged elsewhere raises risk in your file (15 points when flagged elsewhere, 5 when merely present) without exposing any other business's identities, reviews or data. The graph turns isolated rows into patterns that only exist across businesses.

The monthly graph routine, in order

First, your own repeats: a fingerprint with several reviews at your business in a tight window, especially across a burst day, is a local pattern — read its rows together, because together is how they were likely written. Second, flagged-elsewhere appearances: sort by the combination, not the flag alone — an UNLINKED, AI-phrased, burst-day row from a flagged-elsewhere fingerprint is your strongest cluster candidate; a transaction-linked, specific, ordinary row from the same list is probably a genuine traveller who annoyed someone else. Third, note the merely-present count trend: a slow rise is a healthy reviewer ecosystem; a sudden jump alongside other signals belongs in the month's notes. Fifteen minutes, three lists, decisions logged — the graph rewards rhythm, not obsession.

A detail worth knowing before you start: ReviewProof analyses what you give it. The CSV importer reads reviewer, rating, title, body, date and source columns, plus a verified-transaction flag and a transaction reference when you have them. Reviews with a transaction link are marked verified; reviews without one are marked UNLINKED — kept in the audit, visible in the report, but excluded from the verified-experience widget, which only displays reviews that are both transaction-linked and in a kept or published state. That single rule is the difference between a widget that markets and a widget that proves.

What the graph must never be used for

It is not an identification tool: the hash is one-way, the pool carries no identities, and attempting to reverse or triangulate a fingerprint into a person is both futile by design and outside the product's purpose. It is not a public accusation engine: nothing in the graph names a reviewer, and your public replies must never reference fingerprint status ('our system shows you review many businesses') — that is a claim you cannot evidence to the reader and a disclosure of method that helps the actual farms. And it is not a score to optimise: a business that starts refusing service to flagged-elsewhere customers has converted a review-integrity tool into customer screening, a different and uglier product entirely.

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

Treating 'present elsewhere' as guilt — at 5 points it is deliberately a whisper, reflecting how many genuine customers review widely. Exporting graph claims into public replies or marketing ('protected by a cross-business fraud graph') invites challenges you cannot evidence publicly. And checking the graph once, at onboarding, misses its compounding value: each business's honest flags make every other business's next check sharper.

The workflow is the same on every plan, and the free tier does the complete job for one location: import by CSV upload or paste (up to 500 reviews per import on Free), read the full risk analysis and report, open a provenance certificate for any review, and publish the verified-experience widget. Pro ($29/month, 3 locations, 5,000 reviews per import) adds the full provenance chain detail, the evidence export, AI reply drafts and the reviewer-fraud graph. Agency ($79/month, 50,000 reviews per import) is the multi-location tier. Nothing about the free audit is a sample or a teaser — the report you get free is the real analysis of your real rows.

Worked example: the graph catches what the file could not

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

A restaurant's own file looked unremarkable: 233 reviews, risk score 9, no burst. The graph view, however, listed 6 rows whose fingerprints were flagged at other businesses. Reading the combination: all 6 were UNLINKED, 5 carried AI-phrase wording, and their dates — unremarkable inside this file — fell within nine days. The restaurant flagged the 6 internally, logging the combination (flagged-elsewhere status plus links plus wording plus timing), and added its own flags to the pool, which is the reciprocal act that makes the system work for the next business. No single element convicted those rows; the pattern across two layers did, and the log says exactly that.

Your five-action checklist

  1. Run the graph check after every import, on the same monthly rhythm
  2. Work repeats, flagged-elsewhere and merely-present lists in that order
  3. Judge rows on the full combination in your file — never the flag alone
  4. Contribute your own honest flags back to the pool
  5. Never identify, confront or publicly reference a fingerprint

Frequently asked questions

What data does the pooled graph share?

Only the salted fingerprint hash, review counts and flag state with a reason category. Never raw identities, review text, or which merchant flagged it.

How much risk does a flagged-elsewhere fingerprint add?

15 points when flagged at another business; 5 when merely present elsewhere. Both are inputs to your judgement, shown as counts only.

Which plans include the graph?

Pro and Agency. Free workspaces get the full analysis of their own imported file, certificates and the widget.

Can the graph tell me who a reviewer is?

No, by construction. The fingerprint is a one-way salted hash; the system recognises sameness without holding identity.

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.