Your Product Photos Are Your Fingerprint

Dev had been staring at the same photograph for 40 seconds before he worked out what was wrong with it.

It was a kettle. His kettle, matte black, shot three-quarters on a plain grey sweep, with the steam caught at the exact moment he'd spent most of a Saturday waiting for. He knew the frame. He'd chosen it from 180 others.

It was on a listing for 31 dollars, from a seller he'd never heard of, in a currency he hadn't set up. And in the bottom-right corner, half-cropped, was the edge of the apple box his studio assistant had used to prop the handle.

"They didn't even reshoot it," he said, to nobody. "They just took the picture."

Of course they did. Reshooting costs money.

The copy has to show the buyer something

A counterfeiter's problem is the same as yours. Nobody buys a kettle they can't see. So whatever else a copycat listing fakes, it has to put an image in front of the buyer, and it has to be an image of the thing the buyer wants.

They have three options. Shoot the genuine article, which means buying one of yours. Shoot their own version, which means their version has to look as good as yours, and it usually doesn't. Or use your photograph.

They use your photograph. Almost always, in the cheap end of the market, because it's free and it already converts.

That's the fingerprint. Your product description can be paraphrased into something unrecognisable by any decent model in a second. Your brand name can be dropped from the title and buried in the body text. Your photograph can't be meaningfully disguised without destroying the thing that makes it useful to them.

And images are searchable. That's the whole opening.

Four ways to find out who's using yours

None of these are clever. All four take under 10 minutes per image, and most brands have never run any of them.

Google Lens, from the image itself. Upload the file at images.google.com or through Lens and read the visual-match results. This is the one that finds marketplace listings most often, because Lens indexes product images aggressively and marketplaces are well crawled. It also returns a lot of noise, including your own distributors, your own press coverage and your own Pinterest board. Expect to sort.

TinEye. A reverse-image engine that has been doing exactly this since 2008, with one property the others don't advertise: it sorts by first-seen date. If you want to establish that your version of an image existed before someone else's, that sort order is the fastest free evidence you'll find. Its index is smaller than Google's. Run both.

Bing Visual Search. Different crawl, different results, genuinely worth the extra 2 minutes. It surfaces listings on sites that Google has deprioritised, which in this category is often where the copies live.

The marketplaces' own search, by text. Less obvious and more productive than people expect. Search your exact product title in quotes. Search your brand name plus a word you invented. If you named a colourway "Tarn", search Tarn. Invented words are the cheapest tripwire you own, because a copycat scraping your listing copies them without knowing they're a signature.

Run all four on one product and you'll have a count. The count is the point. Not a dashboard, not a programme, a number you didn't have yesterday.

What a match proves, and what it doesn't

Here's where most people get ahead of themselves, and it matters because the next step is a sworn statement.

An image match proves that two listings show the same photograph. It does not prove who took it, who owns it, or that the seller is selling a counterfeit.

Three ordinary explanations come up constantly. The seller is an authorised reseller using your supplied assets, and you forgot. The seller is a dropshipper selling your genuine product, bought from you, at a markup or a discount. The image came from a supplier catalogue that both of you use, in which case neither of you owns it.

That last one bites hardest in categories with shared manufacturing. If your factory shot the product and gave the images to everyone who orders it, your claim to that photograph is weak and your takedown will fail on the first contested notice.

So before anything goes out, you need to answer one question about each image: can I show that I own this? A commissioned shoot with a contract that assigns copyright, yes. A shoot you did yourself, yes. A photograph your agency took under a licence that stayed with them, maybe not. A supplier image, no.

Sort your library into those buckets once. It takes an afternoon and it decides which half of your catalogue you can actually defend. DMCA or trademark? covers what happens next in each case, because an image you can't claim can still be attacked on the brand name in the title.

Build the reference set, then stop thinking about it

The workable version of this isn't a search. It's a short list you maintain.

For each product you care about, pick the two or three images a copycat would steal. In practice that's the hero shot on white and the single best lifestyle frame. Not the detail crops. Not the scale shot with the hand in it. The ones that sell.

Write down, beside each one, the date you first published it and where. A screenshot of your own product page with a visible date, or an email from your photographer with the files attached, is enough to be useful later and costs nothing now.

Then put the official sellers in writing. Your own stores, your distributors, the marketplace shops you run under another name. This list is the thing that turns a noisy match report into a short one, and almost nobody has it written down anywhere a machine can read it.

That's the whole setup. Reference images, first-published dates, allowlist. 90 minutes, once.

Doing it on a schedule without doing it by hand

The manual version works and it doesn't scale past about 6 products, because the searching isn't the cost. The sorting is. Run Lens on 20 images and you'll get a few hundred results, most of which are your own footprint, and you'll stop doing it by week three.

What changes the economics is scoring the results before a human sees them. Photo similarity against your reference image, whether the title repeats your brand or product name, how far below your price the listing sits, and whether the seller is already on your allowlist. Rank by those, dismiss the obvious non-matches, group the rest by seller so 20 listings from one shop arrive as one shop, and the review becomes a 10-minute job instead of an afternoon.

That's what Brand Protection Tools, which Verity AI runs, does with it. Worth saying how, because it's a question we get asked and the answer is unglamorous: it doesn't run its own scrapers. Etsy comes in through Etsy's official API, and the wider web comes through a licensed search API that covers Google web and Google Lens results. The searching is boring and legal. The judgement stays with the brand owner.

It searches Etsy and Alibaba.com now. AliExpress, Temu, eBay and Amazon are next on the list rather than live, which is the sort of thing worth checking on any tool in this category before you pay for it.

The evidence problem nobody mentions until it's too late

Listings disappear. Sometimes because you got them taken down, sometimes because the seller rotated stock, sometimes because the marketplace swept the shop for an unrelated reason.

When a listing disappears, so does your proof. The URL 404s, the images go, and the seller's shop page shows 4 items instead of the 30 it had when you looked. If a counter-notice arrives 2 weeks later and you need to show what was there, a bookmark is worth nothing.

So capture it at the moment you find it, not at the moment you need it. Seller name and seller ID, the listing URL and title, the price, a full-page screenshot, and the date and time you saw it. Put it in a folder named after the seller rather than the listing, because the seller is the unit that matters. The relisting problem explains why that filing choice turns out to be the one that pays off.

Dev found 14 listings using his kettle photograph. 11 were from the same seller under 2 shop names. One was a genuine stockist he'd forgotten about. 2 were from a supplier catalogue and he had no claim on them at all.

He still has the apple box, incidentally. It's in every photo he shoots now, slightly further into frame. He says it makes the matching easier.

For how the same copies end up inside the answers AI assistants give about your brand, start with When the AI shopping assistant recommends the fake.