ReviewProof ⚖

What does duplicate review text tell you?

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

Duplicate text — the same normalised review body appearing more than once in your file — means the words were reused, not independently written. It can point to a copy-paste campaign, a template handed to customers, or staff-written reviews; short generic praise can also coincide honestly. ReviewProof flags duplicated bodies longer than twenty characters and weights the signal at 0.8, so you can open the rows, compare their dates, sources and transaction links, and judge.

What is duplicate review text?

Duplicate review text is two or more reviews whose normalised body — lowercased, punctuation stripped — is identical. Genuine customers describing the same good experience converge on similar sentiments but almost never on identical sentences of any length. Duplication therefore measures something specific: the text had a single origin. Where that origin was (a bought package, a well-meaning template, a staff member) is a second question the surrounding evidence answers.

Length is what makes duplication speak

'Great service!' can recur innocently a dozen times; it carries too little information to prove copying. That is why the detection only counts duplicated bodies longer than twenty normalised characters, and why your reading should weigh length too: three identical eighty-word reviews are a campaign until proven otherwise, while two matching one-liners are noise. Look next at the cluster's shape — same day, same source, handles following a pattern, no transaction links. Duplication plus burst plus UNLINKED is a very different finding from duplication alone, and the report gives you all three in one place.

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.

Templates you hand out count as your handwriting

A common, uncomfortable source is the business itself: a card or message suggesting what customers might say ('mention our fast service!') drifts over months into customers pasting the suggested text whole. The reviews may describe real visits and still be scripted, which misrepresents them. The fix is to strip suggestions back to a bare ask — the request, the link, nothing about content — and to say so explicitly to staff: we never supply wording. The audit then becomes the check: duplicate counts on new imports should fall to the occasional short coincidence and stay there.

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

Confronting a reviewer over a duplicate flag is a mistake — short texts coincide, and the flag is a signal, not an accusation. Deleting one copy of each duplicate to tidy the profile is worse: it edits the pattern out of the evidence. The third mistake is supplying 'example wording' to happy customers to make reviewing easier; you are manufacturing the next duplicate cluster yourself.

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: a repair shop with nine identical paragraphs

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

An auto repair shop imported 231 reviews. The duplicate count was 9 — and opening the rows showed a single 64-word paragraph, word for word, posted under nine handles across eleven days, none with a transaction link. Two further clusters were innocent: pairs of short 'fast, honest, fair price' lines. The owner kept all rows, flagged the nine internally with the pattern logged, and — because the cluster described work the shop could not match to any invoice — treated it as a bought package from a past marketing vendor and documented that conclusion in the record while disputing through the platform. The evidence pack exported the whole picture in one file.

Your five-action checklist

  1. Import your reviews and read the duplicate count in the report
  2. Open each duplicate cluster and compare length, dates, sources and handles
  3. Check the cluster's transaction links — linked rows tell a different story from UNLINKED ones
  4. Remove any suggested wording from your review requests, cards and scripts
  5. Log your conclusion in the moderation record; flag internally, never edit text away

Frequently asked questions

Does duplicate text prove reviews were bought?

No. It proves the text had a single origin. Dates, sources, handles and transaction links around the cluster tell you whether that origin was a campaign, a template, or coincidence on short text.

Why the twenty-character threshold?

Below that length, phrases are too generic to distinguish copying from coincidence. The threshold keeps the signal meaningful: identical sentences, not identical sentiments.

How is duplicate text weighted in the risk score?

At 0.8 per affected row — below gating and incentives, above AI wording — inside the formula 100 × weighted signals ÷ (2.6 × reviews), plus 15 for a burst.

Can real customers write identical reviews?

For a sentence or two of generic praise, yes. For a full paragraph, effectively no — independent writers do not produce identical eighty-word texts by chance.

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.