Product Scout AI
Stop wasting hours on products that flop. Get Winnability Score + PDF report in 30 seconds.
Analyze Product
Analyzing product...
Batch Compare
Compare up to 5 products — one URL per line
Analyzing products...
What Product Scout does
Product Scout reads one AliExpress product listing and returns a single number: a Winnability Score from 0 to 100, plus a WIN, MAYBE or PASS verdict and a short written case for it. You paste a product URL and get a breakdown you can hold the next product up against.
The point is consistency, not magic. Most product research is not difficult, it is slow and uneven. You open ten listings, quietly apply a slightly different standard to each one, and end up picking whichever you happened to look at last. A fixed scoring prompt applied to every product removes that drift.
How the analysis runs
- You paste a product URL. The tool pulls the numeric item ID out of it. Short a.aliexpress.com share links are followed to the real product page first, so those work too.
- It fetches the listing through the AliExpress Affiliate API in USD and English. Four of the fields it reads feed the score: product title, current sale price, positive-feedback rate, and recent order volume. The feedback percentage is converted into a 0 to 5 rating. If the item cannot be read — a dead or private link, or an API error — you get a notice saying so, not a score.
- Those four fields go to a language model with a fixed scoring prompt, along with a review count that is always zero because the API is never asked for one. The prompt tells the model to score five factors from 0 to 100 and to return their average as the Winnability Score; the number that comes back is displayed as-is.
- The prompt sets the verdict bands — 70 and above is WIN, 40 to 69 is MAYBE, anything under 40 is PASS — and the model returns the verdict with a summary, a recommended selling price, risk flags and strengths.
- The finished report is stored against your account. Downloading the PDF re-renders that saved report instead of running a fresh analysis, so the number in the file always matches the number you saw on screen.
The five factors
- Demand is the model's read of how wanted the item is. Recent order volume is the only demand figure it is given.
- Margin is judged from the wholesale price. The prompt names $15 to $50 as the optimal band.
- Competition is the model's estimate of how crowded the item looks. A high number means more sellers, not a better product.
- Trend is whether the item reads as still climbing or already past its peak.
- Supplier comes from the seller's positive-feedback rate.
What you get back
On screen: the score, the verdict, the summary sentence, a bar for each of the five factors, a recommended selling price, and badge lists of risk flags and strengths. The PDF is three pages, a cover with the score, the analysis with factor bars, and the raw product data the score was built from. Rather than invent a specimen product, here are the fields a result is actually made of:
The Batch Compare tab takes two to five URLs, one per line, scores each and sorts them highest first in one table. That is the mode worth using. A single score is hard to read; a score sitting next to four others is not.
What this will not do
It only reads AliExpress. Temu, Amazon, Shopify, 1688 and Alibaba links come back with a notice rather than a fabricated number, because there is no reliable read for them here yet.
The score is built from four listing fields. It does not touch Google Trends, ad libraries, competitor store counts, sales history, or your own store data. Whatever it says about trend and competition is inference from a product title and an order count, not measurement. Those two factors are the softest numbers on the page and should be treated that way.
The rating is derived from the seller's positive-feedback percentage, not a star average, and the tool never asks the API for a review count — it sends zero — so review text and review volume play no part in the analysis. Running the same URL twice can return a different total, because the model is not deterministic.
The score is whatever the model returns. Nothing here recomputes the average from the five factors or checks the verdict against the score, so the two can disagree. If the model call fails outright, or its reply cannot be read as JSON, the tool falls back to a rule-based scorer built from the same price, rating and order numbers; that fallback uses a different formula, so a score produced by it is not directly comparable to one from the model.
It does not check trademarks or patents, shipping times to your market, whether the supplier replies to messages, or whether the product will survive an ad platform review. It does not know your niche, your audience or your budget. Use it to order a shortlist and to catch obvious problems early. Deciding to spend money is still your job, and it should still involve reading the one-star reviews yourself.
Where it fits with the other tools
Scout is the first pass. When a product survives it, put the shortlist through the quick winnability check in Product Validator for a second opinion before any budget moves, then write the store copy with the SEO product listing generator. If you are selling outside the English-speaking market, run that finished copy through cross-border localization instead of leaving it to a machine translation at checkout.
Frequently asked questions
Is Product Scout free?
Yes. Every analysis is free, there is no card and no per-analysis cap. You do need a free account, because each report is saved against it and the PDF download re-renders that saved report instead of running a second analysis.
Do I need an account to use it?
Yes. The analyse button opens a sign-up or login prompt on the page itself, and analysis only starts once you are signed in. Registration is free and there is no daily quota once you are in.
Which stores can I paste links from?
AliExpress product links only, including short a.aliexpress.com share links, which are followed to the real item page. Temu, Amazon, Shopify, 1688 and Alibaba URLs are rejected with a notice rather than scored on guessed data, because there is no reliable read for them yet.
What data does the Winnability Score actually use?
Four fields from the listing: product title, current sale price, the seller's positive-feedback rate converted to a 0 to 5 rating, and recent order volume. It does not pull Google Trends, ad libraries, competitor store counts or sales history. Everything the score says is an inference from those four numbers.
How accurate is the score?
It is a structured second opinion, not a forecast. Demand, margin and supplier are anchored to numbers taken from the listing — order volume, price, feedback rate — while trend and competition are the model's inference from a product title and an order count, so they are the softest numbers on the page. Nothing recomputes or checks the model's arithmetic, and running the same URL twice can return a different total, because the model is not deterministic.
Can I compare several products at once?
Yes. The Batch Compare tab takes two to five URLs, one per line, scores each one and sorts them highest first in a table with score, verdict, price, demand and margin. Links it cannot read are dropped from the table rather than scored. Comparing a shortlist is more useful than reading a single score in isolation.