How it works

Unlike other fake-review tools, we publish every criterion. We don't use review text — only the structural data on the public page — to compute a trust score.

Signals we use (no review text needed)

Unnatural star distribution — an excess of 5-stars (roughly 92%+), or a bimodal 5★/1★ split, suggests manipulation.
Number of ratings — high averages on few ratings are unstable; bigger samples are more reliable.
Verified-purchase share — the portion of shown reviews marked "Verified Purchase". Fakes trend low.
Posting bursts — shown reviews clustered on a single day suggest a review campaign.
Average vs distribution — if the implied average from the distribution disagrees with the shown average, data confidence drops.

"Recommended" requires all three

① High trust score — low fake-review signals (trust score 75+).
② Genuinely well-rated — real reviews but a low average means a mediocre product; we don't push it (★4.0+ guideline).
③ Enough data — distribution, rating count, and shown reviews are present, so quality can be confirmed (confidence Medium or High).

Only products meeting all three are recommended with an Amazon link.

We only feature the good ones

Products that don't qualify are not disparaged — simply not listed (with the reason shown on each). Verdicts are estimates from public structural data and don't guarantee authenticity; mistakes are possible. Make the final call yourself.

Try it now

Paste an Amazon product URL and we apply exactly this method to score fake-review risk.

See our fake-free picks → How to spot fake reviews →

FAQ

Q. What are the three conditions for a "Recommended" verdict?

A trust score of 75 or higher (low fake-review signals), an average rating of at least 4.0 (genuinely well-rated), and enough data to judge confidently (star distribution, rating count, and shown reviews present). Only products meeting all three are recommended.

Q. What happens to products that don't qualify?

They are not disparaged — simply not listed, with the reason shown on each.

Q. Do you use the review text?

No. We analyze only public structural data — star distribution, rating count, verified-purchase share, and posting-date clustering — and never store or republish review text.