Product Validator
A reality check on one AliExpress link before you spend a dime
No credit card needed · Unlimited after free sign-up · AI-powered by DeepSeek
What the Product Validator does
The Product Validator is a pre-flight check on a single product. You paste one AliExpress product URL and get back a Winnability Score out of 100, a WIN / MAYBE / PASS verdict, five factor scores, a short list of strengths and risk flags (the prompt asks the model for at most three of each), a margin estimate and a suggested platform to sell on. It is the cheap step you run before the expensive one. Killing a bad idea here costs nothing; killing it after an ad test costs whatever you put into the ads.
It is deliberately narrow. One product at a time, one screen of output, and an account that needs nothing but an email address, a password and a click on the confirmation link. It does not build a store, order samples, contact a supplier or write your ads.
How it works
- You paste a product URL. The product name field is optional and is usually ignored: when the listing is read successfully the model sees the title that came back with the record, and your text is only passed along if the lookup itself errors out.
- The server checks that you are signed in. Each validation calls a paid model, so anonymous requests are refused before any money is spent rather than after.
- The URL is resolved to the record for that item: title, current sale price in USD, buyer rating on a 5-point scale, and recent order volume. That lookup goes through the official AliExpress Affiliate API when API credentials are configured, and falls back to reading the listing page directly when they are not. A link that is not AliExpress, or an item that cannot be read, stops here with a notice instead of a score.
- Those fields are sent to DeepSeek with a fixed scoring prompt. The model rates five factors from 0 to 100: demand, margin potential (the prompt treats a $10-$50 spread as the useful range for dropshipping), competition inverted so that a higher number means a less crowded niche, trend, and trust signals around the product and its supplier.
- The same prompt asks the model to average those five factors into the Winnability Score and to apply fixed cuts to it: 70 and above is WIN, 40 to 69 is MAYBE, below 40 is PASS. The score and the verdict are read straight out of the model's reply — the server does not recompute the average or re-check the verdict against it.
- The model also returns a one-sentence summary, the strengths and risk flags, a dollar margin range and a recommended platform. The page draws the factors as score bars so you can see which one is pulling the score down.
What you get back
The page lays every result out the same way. Below is an illustrative example, not a real user's report, to show what the fields look like when the model fills them all in. Anything the model leaves out is shown as "N/A" or "None identified" rather than filled in for it.
The competition score is the one people misread. It is inverted on purpose, so 38 means crowded, not calm. When a score looks strange, open the factor bars before you argue with the verdict.
When it helps and when it does not
It earns its keep when you are staring at a browser full of supplier tabs and need to cut the list down before you invest real attention. It is fast enough to run on everything you saved this week and specific enough to tell you why something is weak.
AliExpress only. Other stores are refused rather than guessed at. If you validate products from Temu or a competitor's Shopify page, this tool cannot help you yet.
The score is a model's judgment, not a measurement. It reads four data points about the item and reasons from general knowledge. It does not query search volume, ad costs, trend charts or competitor storefronts, so it cannot tell you a trend peaked six weeks ago. Run the same product twice and the score can move a few points.
No trademark, patent or ad-policy check. Nothing here queries a trademark register or a platform policy database. A branded or restricted item can still score well. That check is still yours to do.
Trust is inferred, not audited. The trust factor is the model's read of the rating and order count on the listing, not the result of any supplier verification. Nothing here vets a supplier. It says nothing about shipping times or how returns will actually go.
It knows nothing about you. Your creative, your audience and your cost per click decide most of the outcome, and none of that is in the input. The same product can be a 75 for an operator with a working hook and a loss for everyone else.
If the model call fails you get a placeholder. A result of exactly 55/100 with generic wording means the AI request did not go through, or came back without usable JSON, and a canned response was shown instead. Run it again rather than acting on it.
Where it fits with the other tools
Most people use this first and go deeper on whatever survives: run the shortlist through the Product Scout for the longer analysis and a PDF report, use the Supplier Finder when you want a local supplier in the UK, DACH or Australia instead of a long shipping window, and once you have committed to an item, generate the store copy with Perfect Listing. All of them are free with the same account.
Frequently asked questions
Is the Product Validator free?
Yes, and there is no credit card step. You do need a free account, because every check calls a paid AI model and the server rejects anonymous requests before spending anything. Once you are signed in there is no cap on how many products you validate.
Which product links can it read?
AliExpress product URLs only. The listing is read through the official AliExpress Affiliate API when API credentials are configured, and by reading the listing page directly when they are not. Links from Temu, Amazon, Alibaba, 1688 or a Shopify store return a plain notice instead of a score. That is deliberate: without real product data the model would be scoring a URL string and inventing the numbers.
What data does the score actually use?
Four fields from the AliExpress record for that item: title, current sale price, buyer rating and recent order volume. Everything else in the output comes from the language model reasoning over those four fields. There is no search volume feed, no ad cost data and no view of competitor stores.
How accurate is the Winnability Score?
It is a model's opinion, not a measurement, so treat it as directional. Two runs on the same product can land a few points apart. It is useful for comparing a shortlist and for catching obvious problems, and it is not a substitute for reading the worst reviews and doing your own margin math on the products you are serious about.
What do WIN, MAYBE and PASS mean?
The prompt asks the model to average the five factor scores and to apply fixed cuts: 70 and above is WIN, 40 to 69 is MAYBE, below 40 is PASS. Both the score and the verdict come back from the model as written — the server does not recompute the average or re-check the verdict against it — so read the factor bars next to the badge rather than the badge alone.
How is this different from Product Scout?
The Validator answers one question about one link, so you can run it on a handful of candidates while you are still browsing. Product Scout does the longer analysis on the products that survive and gives you a PDF report you can keep.