I checked 29,529 AliExpress discount claims against my own price data. 21,720 were inflated.
I run PriceProven. It records the price of AliExpress listings once a day and keeps the dated readings, so a "50% off" badge can be checked against what the price actually was.
Where the numbers stand today (17 Sep 2026): 81,650 listings tracked, 1,231,139 price readings taken, 43 days of our own history. Of 29,529 seller discount claims we have checked, 21,720 did not survive the check.
The pattern is duller than "sellers are lying". Mostly the crossed-out price is a number the listing has never actually charged in the window we have watched. One example from today: a car star-projector shows "59% off $4.83". In 40 days of readings the highest we ever recorded was $2.32. It is at $1.97 now, which really is its lowest with us. The discount is real; the reference price is fiction.
What this method cannot do, so nobody over-reads it: we only know a listing from the day we started watching it, so a genuine sale that happened before then is invisible to us. We compare a listing to its own past, not to other shops selling the same thing. We do not verify quality, shipping or seller behaviour. Listings without enough history get labelled as such instead of getting a verdict.
The thing I got wrong early on: I assumed the interesting output was "this discount is fake". It isn't. Nobody wants to be told off for wanting a deal. The output people actually use is "this is above its own usual price right now, here are cheaper listings we track" - a decision, not an accusation. Rewriting around that changed the product more than any feature did.
For anyone doing data-backed content: the hardest part was not collecting the data, it was resisting rounded-up claims. Every time I wrote "most discounts are fake" I had to replace it with the counted figure and the date, because the figure moves daily. Duller copy, but it is the only thing that survives someone checking.
Question for this crowd: for a trust-first tool, how early did you put an email wall in front of your data? Mine gates the price history and I cannot tell yet whether it is collecting the right people or just costing me the curious ones.
I reckon the price history should stay open, especially when trust is the product. if i need to give you my email before seeing the evidence, i can’t tell yet whether the tool deserves that trust.
I'd gate the ongoing value instead — price alerts, watchlists or weekly updates. let people check a couple of listings freely, then ask for their email when there’s a clear reason to return. probably fewer signups, but more of the right ones.
The "insufficient history" bucket is the part I'd watch closest. That's where your confidence is lowest and it's tempting to just flag it and move on, until someone hits a listing that's a genuine steal but got lumped in with the noise because it's new. Might be worth surfacing days-of-history as its own number next to the verdict instead of a binary label, so people can judge how much to trust it instead of you deciding for them. Also worth knowing: the anchoring-price problem you're describing isn't just an AliExpress thing, the FTC went after Wayfair and a few others for exactly this (crossed-out price that was never a real price), so there's real precedent behind the pattern you're finding. Could be a good line for your credibility copy.