How Detection Works

Education

The Rating-Text Mismatch Signal: When the Stars and the Words Disagree

Rating-text mismatch flags a five-star review with no real detail, or a one-star review with no actual complaint in it. It's the lightest of our four signals alone, but it sharpens every other flag it appears next to.

Andrew Keymaster · Sep 18, 2026 · 7 min

Education

Duplicate Text and Generic Accounts: The Rarest (and Most Telling) Fake-Review Signals

Duplicate text and generic accounts are the two rarest signals in our fake-review scoring model, flagging just 283 and 526 reviews out of 92,359 flagged as of September 2026. They're also the closest thing to catching manipulation in the act, rather than inferring it from timing alone.

Andrew Keymaster · Sep 13, 2026 · 6 min

Education

The Low-Activity Account Signal: Our Most Common Fake-Review Flag, Explained

Low-activity reviewer accounts, profiles with no other reviews, photos, or history, make up 91% of everything Revdict flags. Here is why the signal is so common, why it is not proof by itself, and how it combines with other patterns to move a verdict.

Andrew Keymaster · Sep 4, 2026 · 6 min

Education

What Is a 'Review Drive'? The Manipulation Pattern Where the Customers Are Real

Our scoring model flags a distinct pattern separate from fake or bought reviews: staff prompting real customers, in a compressed burst, to leave five-star reviews naming a specific employee. Here's how it's detected, how often it shows up, and why Google banned the practice in April 2026.

Andrew Keymaster · Aug 31, 2026 · 6 min

Education

The Economics of Review Farms: What Review-Buying Actually Costs

Fake five-star reviews sell for as little as $9 each; fake negative reviews aimed at a competitor cost more. Here's what the review-manipulation market actually charges, who profits, and what a business risks under the FTC's $53,088-per-violation rule.

Andrew Keymaster · Aug 8, 2026 · 5 min

Education

Review Bursts: The Clearest Fake-Review Signal, Explained

A review burst is a sudden spike of five-star reviews arriving far faster than a business's normal pace — and it's the single strongest signal in our fake-review model. Here's what counts as a burst, why timing matters more than content, and what real burst patterns look like in our data.

Andrew Keymaster · Aug 6, 2026 · 6 min

Education

How Our Fake-Review Score Works (and What It Can't Tell You)

Our 0-100 manipulation score combines review-burst detection, reviewer-account age and activity, duplicate-text matching, and rating-text consistency into a single number. Here's what goes into it, and the real limits of what any automated score can prove.

Andrew Keymaster · Jul 21, 2026 · 6 min