No install · Free · Runs in your browser
We measure what the star rating hides — AI-written text, reworded copies, undisclosed sponsorships. You see which sentence was flagged, and why.
Input
The quality is outstanding. Delivery was prompt and packaging was perfect. Overall a truly great purchase.
Signal readout
No install, no signup. Paste reviews on this page and get results instantly.
Use nowOpen a product page and each review is scored automatically, right where you read it.
InstallSave it to your bookmarks bar once, then click it on any shop to check that page.
Save itWeb checker
The same engine the extension uses. What you paste never leaves this browser.
Nothing sent · nothing stored · no account
When copying is blocked
Many sites block text selection on reviews. Here is what to do instead — start from the top.
No copying needed at all. Save it once to your bookmarks bar, then click it on any page with reviews to measure them in place.
The extension skips copying entirely. Open a product page and scores appear next to each review.
If long-press selection does not work, use your browser's Share or Copy page text. Menus, prices and dates will come along — the checker filters those out for you.
Two or three suspicious ones are enough. Enter similar-looking reviews together and we cross-check them for copy-paste.
Copy protection is the site's own policy. This service does not circumvent it — it simply reads the text already displayed on your screen.
Measured accuracy
Our first version caught zero fakes. We publish that failure with every other number.
This tool is for people who distrust reviews. Hiding our accuracy would defeat its purpose.
| Engine | Ordinary fakes | Deliberate evasion |
|---|---|---|
| v1 (initial) | 0% | — |
| v2 | 100% | 31% |
| v3 (current) | 100% | 33% |
Why trust this
Measurement finishes in your browser. Only a handful of numbers are ever sent, and the original text cannot be reconstructed from them.
We do not just say "suspicious". We show how many praise phrases, whether there is any real-use detail, and what percentage overlaps another review.
These are reference indicators. We do not declare anything fake — the final call is yours.
Examples
Every score below came from the real engine. The last card is one it missed, shown as is.
Excellent product! Fast shipping and great quality. Highly recommend. Will buy again.
High suspicion — praise phrases 4 only, no real-use evidence
The most common type. Nothing but praise, no sign of what was actually used.
The quality is outstanding. Delivery was prompt and packaging was perfect. Truly satisfied.
High suspicion — praise phrases 3 only, no real-use evidence · formal throughout with no usage detail
Dressed up in formal language, yet still zero real-use description.
Fast shipping thank you will use well
High suspicion — overlaps another review at 100% match — reworded copy
Short, formulaic thanks. Ambiguous alone — but compare it with the next card.
Quick delivery thanks a lot will use it well
High suspicion — overlaps another review at 100% match — reworded copy
"Fast → Quick", "shipping → delivery". The sentence skeleton is identical.
Must buy!! 🔥 Pros: amazing value 👍 Cons: none 💯 #ad #musthave #bestbuy
High suspicion — sponsorship disclosed · promotional clichés 2
Emoji, hashtags and pros/cons formatting — advertising style with no disclosure.
Been using it 3 weeks. The motor is louder than I expected, kind of annoying at night.
Looks genuine — real-use traces 2 (parts, sensory) · complaints/reservations 2 — typical of a real user
Duration, a sensory complaint, hedging. You cannot write this without using the thing.
Purchased this for my father. The buttons are somewhat small, however the operation is reliable. Weight is approximately 2kg.
Looks genuine — real-use traces 2 (parts, context) · complaints/reservations 2 — typical of a real user
Formal, yet it names buttons and weight. We judge content, not writing style.
The button presses smoothly and there is no noise at all. No heat, very satisfied.
Looks genuine — real-use traces 4 (parts, sensory)
Sensory words faked in a purely positive direction. Scored 84 and passed. We still miss these.
FAQ
Blocklists of "fake phrases" break the moment someone swaps a word. So we do the opposite: we look for traces only a real user leaves — part names, sensory detail, genuine complaints, situational context — plus structural overlap with other reviews.
No. Measurement finishes inside your browser. Only a few numbers used for scoring are ever transmitted, and the original text cannot be recovered from them.
Mobile Chrome does not support extensions. Paste reviews into the web checker above instead — it runs the exact same engine.
Save the bookmarklet and it measures the page directly, no copying required. You can also use your browser's "copy page text" — menus, prices and dates get filtered out automatically.
We catch most ordinary fakes, but only about a third of text written deliberately to evade our criteria. That number is published above. Results are reference indicators, not verdicts — which is why we always show the reasoning.
No account, no payment. Mobile: web checker. Desktop: extension.
Last updated: August 14, 2026
By default, nothing. All analysis happens locally in your browser.
Anonymous statistics may be collected only if you explicitly opt in:
These statistics contain no review text, no product names, no user identifiers, and no IP addresses.
This website records anonymous visit statistics to improve the service.
If a brand or manufacturer uses the audit report API separately, submitted review text is scored immediately and only the score is stored — the original text is not retained (identified by SHA-256 hash only). Email addresses provided for API access are used solely to provide the service.
Opt-in anonymous statistics are stored only in aggregate form for service improvement and are not linked to individual sessions. Score data from B2B audit reports is deleted immediately upon request.
The browser extension requests the following permissions:
Privacy inquiries: [email protected]
If this policy changes, the revised version and its effective date will be posted on this page.
Last updated: August 14, 2026
Review Assay (the "Service") measures specific signals in shopping review text and displays the result along with the reasoning behind it.
The Service is provided as-is. The operator makes no warranty as to the accuracy or completeness of results, and to the extent permitted by law is not liable for direct or indirect damages arising from use of the Service.
The shopping platforms analyzed have no affiliation with the operator. Their trademarks belong to their respective owners.
The Service may display advertising to cover operating costs. Advertising is shown clearly separated from the analysis area, and advertisers have no involvement in any verdict. The operator is not responsible for the content of external sites reached through advertising.
The operator may change or discontinue the Service. No compensation is provided for changes to or discontinuation of a service offered free of charge.
These terms are governed by the laws of the Republic of Korea. Disputes relating to use of the Service will be resolved in the court having jurisdiction under the Korean Civil Procedure Act.
If these terms are revised, the changes and their effective date will be posted on this page. Continued use after revision constitutes acceptance of the revised terms.