How the A–F trust grade works
The ShopSmart-AI trust grade is a single A–F letter built from four groups of signals on a listing. Those four groups are the listing itself, review authenticity, seller and brand reputation, and counterfeit indicators. A grade is only worth something if you can see what is behind it. Below is everything we check, and what the grade does not claim to do.
What we read off the listing
Every scan starts by reading the product page itself — title, description, images, specs, price, the seller block, and the review text the page exposes. A Deep Scan pulls a larger sample of reviews than a quick scan.
- Product title, description, and spec claims
- Listed price against the going rate for the same item
- Seller name, storefront, and fulfilment details
- Review text, ratings distribution, and how recent the reviews are
Review-authenticity signals
This is the fake-review part. We look for the patterns that show up when reviews have been bought, incentivised, or generated rather than written by buyers.
- Bursts of reviews clustered in a short window
- Repeated phrasing and near-identical wording across reviewers
- Reviews that read as AI-generated rather than written by a buyer
- Praise that does not match what the product actually is, or contradicts the complaints
- Ratings distribution that skews to 5-star and 1-star with nothing in between
Seller and brand reputation
A clean listing on a bad seller is still a bad buy, so we search the open web for what other people say about the seller and the brand — outside the marketplace, where the seller does not control the page.
- Live web search results for the seller name and the brand name
- Complaint patterns: non-delivery, wrong item, refused returns, warranty dodging
- How long the seller and brand have been visible, and under how many names
- Whether the brand exists anywhere outside this one marketplace listing
Counterfeit indicators
Counterfeits usually leave a trail on the listing before they leave one in the reviews.
- Price well below the real market price for a branded item
- Brand name that is a near-miss on a known brand, or a generic brand on a branded product
- Mismatched or reused product images, or images that contradict the spec text
- Buyers reporting packaging, serial numbers, or build quality that is off
How it becomes one grade
The evidence above goes to an AI model that weighs it and returns three A–F grades — product, seller, and brand — plus the red and green flags that drove each one, and a one-paragraph plain-language summary. You always see the flags, so you can disagree with the grade.
- A — nothing concerning found across listing, reviews, and reputation
- B–C — mixed: real flags exist, read them before you buy
- D–F — clear risk signals; the flags explain which ones
- N/A — the page is not a product page, or there was not enough evidence to grade
What this grade is not
- Not a safety certification. We do not test products, inspect them, or verify regulatory compliance.
- Not a legal judgment. A low grade is not an accusation of fraud against any seller or brand — it is a read on the evidence a careful shopper could find themselves.
- Not a guarantee. An A-graded seller can still ship you a bad unit, and an F-graded one can be fine.
- Not a price checker. We look at price only where it is evidence of a counterfeit, not to find you a deal.
- Not human-reviewed. Every grade is produced by an AI model reading public evidence. It is a research shortcut, not a verdict.
- Not sponsored. No seller, brand, or marketplace pays to change a grade, and we do not sell your data.
Questions this does not answer? Try the FAQ, or see what a scan costs.