The listing that is not a seller: scraper storefronts and ghost inventory
How a storefront with no stock takes your order
A shopper searches for a popular item and finds two listings at nearly the same price. One belongs to a seller with a warehouse. The other belongs to nobody. It was built by a scraper that copied the title, the photos, the description, and the reviews of the real listing, then priced it a few cents lower so the algorithm puts it first.
When an order comes in, the operator has two choices. The honest version of the scam buys the item from the real seller at full price and ships it to the buyer, pocketing the difference between the two prices, which is usually nothing, so this version rarely lasts. The common version takes the payment and ships nothing, or ships a counterfeit, or ships late from a different supplier while the buyer waits past the refund window.
DataDome's 2026 State of Bot and Agent Security report puts scraping at 70.9 percent of all bad bot traffic in the year to June 2026, up 185.2 percent year over year. Price scraping is the largest single use of that traffic, because a price is the most valuable fact a competitor can steal and the easiest to collect at scale. A scraper storefront is what that stolen data becomes when it is turned into a listing.
Why the marketplace struggles to catch them
The storefront looks like a seller. It has a name, a rating, and a catalog. The rating is often real, in a sense: the operator buys a few cheap items from real sellers, leaves five-star reviews for themselves through other accounts, or simply inherits the review count of the copied listing in the shopper's mind. Shoppers do not audit a storefront. They compare prices.
The inventory is the tell, but only if you watch it move. A real seller's stock falls when orders arrive and rises when a shipment lands. A scraper storefront's stock never falls, because there is no stock, and it never runs out, because the listing is regenerated from the scrape. A listing that has been in stock at the same quantity for three months, across a hundred products, is not a warehouse. It is a database.
The operator also rotates. When a storefront gets flagged, the listings reappear under a new name within hours, because the catalog is a scrape and the scrape is automated. Banning the storefront is banning a printout. The scraper prints another one.
What to check before you buy from a storefront you do not know
Read the seller's history, not the listing. A seller who has sold a hundred different products for two years is a business. A seller who appeared last month with a thousand listings across unrelated categories is a catalog. Nobody stocks phone cases, power tools, and baby clothes from the same shelf.
Check the review pattern. Reviews clustered in a single week, all five stars, all short, all for different products, point to manufactured history. Real sellers accumulate reviews slowly, unevenly, with the occasional complaint about shipping.
And check the price against the pattern. A listing a few cents below every competitor, on every product, at every hour, is priced by a rule, not a person. A person reprices in batches during working hours and misses changes. A bot reprices constantly and misses nothing.
What marketplaces should do about it
The fix is inventory proof. A listing should have to show where the stock is, at least in aggregate: a warehouse location, a supplier invoice, a shipment record. Marketplaces already ask for this when a seller scales up. The gap is at the bottom, where a new account can list a thousand items on day one with no evidence of a single unit.
Velocity limits on new accounts are the second control. A storefront that lists a thousand products in its first week is not a new business finding its feet. It is a scrape being published. A ramp, a hundred listings in week one, a thousand after a month of clean sales, costs real sellers nothing and costs the operator the speed the model depends on.
The third control is the simplest: when a buyer reports a storefront that never shipped, the marketplace should check the seller's other listings for the same pattern instead of refunding one order and moving on. Ghost inventory is a catalog problem, and catalogs are visible. One complaint should trigger a catalog review.
Keep reading: How do bots harvest seller phone numbers from marketplace listings?, Why do bots scrape your prices and undercut you by a cent?, and How do bots use marketplace messages to phish sellers out of their accounts? cover the nearby cases in this series.
Frequently asked questions
How can I tell if a marketplace listing is a scraper storefront?
Look at the seller, not the price. A storefront that appeared recently with hundreds of listings across unrelated categories, reviews clustered in one week, and stock that never runs out is usually a copied catalog. A real seller has a history in one category and stock that moves.
What happens to my money if I order from a ghost listing?
If the operator never ships, you can usually get a refund through the marketplace's buyer protection, but it takes days or weeks, and the refund window can close if the operator strings you along with fake tracking. Report the listing as soon as the delivery date passes, not after a second promised date.
Why do marketplaces allow new sellers to list thousands of items?
Because friction at signup drives away real sellers too, and most new accounts are legitimate. The tradeoff is real, but the fix is a ramp: let new sellers list freely in small numbers and require proof of stock before the catalog scales into the thousands.
Is scraping prices illegal?
Scraping publicly visible prices is usually not a crime on its own, but using the scraped data to mislead shoppers, copying listing text and images, or breaching a site's terms can be. The practical defence is technical and commercial: make the scrape expensive and the fake listing unprofitable.