That flawless product photo, the lamp glowing in a perfect living room, the handmade vase with not a speck out of place, may never have been photographed at all. AI-generated product photos are listing images made by an image generator instead of a camera, and they are now common on Amazon, Etsy, eBay and Facebook Marketplace. This guide shows you how to check a listing photo before you pay, using the same checks that catch stolen catalogue shots and dropshipped goods.
Direct answer: AI-generated product photos are listing images made by an image generator instead of a camera, and you can usually catch them by combining three checks: reverse-search the photo, compare every photo in the set against the others, and read the buyer photos. No single tell proves a photo is fake, so judge the pattern, not one detail.
An AI-generated product photo is a picture of an item that was never placed in front of a camera. The generator invents the product, the setting and the lighting in one pass, so the result can look more polished than a real studio shot. That polish is the trap: shoppers trust photographs, and a generated image lets a seller advertise an item they do not own, have never held, or cannot supply in the condition shown. AI-generated product photos are not a rare edge case in online shopping any more, in September 2026 the BBC’s Rip Off Britain examined Facebook Marketplace listings and found AI-generated images being used for items that did not exist.
This guide covers product imagery on platform listings only: whether the photo itself shows the real item. It does not judge whether an unknown standalone shop is legitimate, that is a different question, covered in our guide to checking whether an online shop is real. And it is about products, not people: our guide to fake before-and-after photos covers transformation photos of people and results, while this one covers photos of things for sale.
What is an AI-generated product photo?
An AI-generated product photo is any listing image in which a generative image model created the item, the scene, or both, instead of a camera recording something real. Three kinds turn up most often, and telling them apart helps you choose the right check.
- Fully generated. The whole image, product, background, shadows, props, was invented by a model. Nothing in the frame was ever photographed. These are the listings where the item itself may not exist.
- Real product, invented scene. A genuine photograph of the item is composited into a generated room, garden or studio backdrop. The product may be real, but the setting, scale and lighting are invented, so the photo still misleads you about size and finish.
- Embellished real photo. A real photograph is heavily edited by AI: colours deepened, scratches removed, proportions tidied, a worn surface made to look new. The item exists, but not in the condition the photo promises.
Not every generated image is a scam. A clearly labelled concept render for a pre-order, or a real product placed on an AI-made background with the product itself unchanged, can be honest presentation, AnyFact’s marketplace guide draws exactly that line, treating a generated image shown as the real item as the case that is not allowed. The problem is the undisclosed kind: the photo presented as a straight photograph of the item you will receive.
AI product images on Amazon and Etsy spread for a simple reason: a generated set costs nothing and looks better than honest photography. A new seller account can list dozens of items in an afternoon without owning a single one, and the photos are the only thing the buyer can judge. That is also why the photo is the weak point, the same techniques that expose a stolen catalogue photo expose a generated one, as long as you check before you pay. The US Federal Trade Commission warns, in its online shopping guidance, that if expensive brand-name items are offered for bargain prices, they could be counterfeit or stolen, and a too-good price riding alongside a too-perfect photo is worth a minute of doubt.
How can you spot AI-generated product photos?
Learning how to spot AI generated product images starts with zooming in, not stepping back. Generators are good at the overall impression and weak at the small consistencies a camera gets for free. Work through the tells below as a set, and remember the rule this whole guide runs on: judge the pattern, never a single clue.
- Text and labels melt. Zoom into any printed text, the brand name, the care label, the writing on the box. Generated text often comes out warped, misspelled, or as nonsense characters that almost look like letters. Real photography does not do that.
- Hands and fingers bend wrong. A hand holding the product is a classic giveaway: fingers that bend the wrong way, too many of them, or a grip that could not hold the object. The same applies to straps, handles and clasps that pass through the hand instead of around it.
- Details drift inside one image. Check the repeating parts: the buttons on a shirt, the links of a chain, the slats of a chair. A generator often changes the count or shape of a repeated detail halfway across the frame, while a camera keeps it consistent.
- Reflections and shadows disagree. Look at glossy surfaces and ask whether the reflection matches the room, and whether all the shadows fall in the same direction. Two light sources in one image, or a shadow pointing the wrong way, is a strong sign the scene was assembled.
- Everything is too perfect. Textures that look dipped in wax, fabric with no crease where a crease should be, a “handmade” item with no trace of a maker’s hand. Real objects have small imperfections; generated ones often have none at all. On clothing this matters most: a generated photo can show fabric hanging perfectly while the real garment fits and drapes differently, so a flawless listing photo is still worth checking against buyer photos.
- Backgrounds smear or repeat. Blur that turns into a repeating smear, tiles or bricks that change pattern mid-wall, plants that dissolve into the wall behind them. The background is where generators run out of attention first.
Here is the honest caveat, and it matters more than any single tell. The newest image models have mostly fixed hands and text, so a clean image is no longer proof of a real photo, that is the warning in FakeImageDetector’s guide to spotting fake product photos, and it is why this guide never rests a verdict on one detail. A clean photo is not proof either way. It may be a genuine photograph, or it may be a generated image from a model that has learned to spell. That is why the visual tells are only step one: they raise suspicion, and the reverse search and the set comparison below are what settle it.
It is also worth saying plainly what a tell does not mean. One warped label does not prove a seller is dishonest, and it is certainly not an accusation to post in a review. A small seller may use AI backgrounds on a genuine item, or a marketplace may compress a real photo into something strange-looking. The checks in this guide are for your own decision before you pay, not a verdict on a real seller, and never a reason to accuse one in public on the strength of a single detail.
How do you reverse-search a listing photo before you buy?
A reverse image search is the fastest check you have, and it catches the most common fakes: stolen or stock photos, and shots reused across many sellers. Run it like this:
- Save the largest version of the main photo. On a computer, right-click and save the image; on a phone, long-press and save it, or screenshot the listing and crop to the product. A thumbnail gives the search engines too little to work with.
- Upload it to Google Lens. Lens lets you search with a camera, an image or a screenshot, and its shopping results show who makes an item, how much it costs and where to buy it. Try TinEye and Yandex as a second opinion, since each engine crawls slightly different corners of the web.
- Read the pattern of results, not just the first match. If the exact photo appears on the brand’s official site while you are buying from an unfamiliar seller, the seller almost certainly does not have the real product in hand. If it appears in a stock photo library, it is a generic image, not a picture of what will ship.
- Repeat with a second photo from the same listing. One search can miss; two searches from different angles rarely do. Our guide to search by image for shopping walks through the same technique for finding the cheapest genuine listing of an item.
- Note what does not come back. A brand-new AI-generated image often has no earlier copy anywhere on the web, so a search that returns nothing is not a pass, it is an unanswered question, and the set comparison in the next section is what answers it.
Two result patterns deserve a closer look. First, the same photograph used by a dozen unrelated sellers at wildly different prices: that is the dropshipping tell, and it means nobody in the chain has photographed the real item. Second, a photo that turns up on the brand’s own website or a major retailer while the listing claims to be the original seller, the listing is using someone else’s catalogue. Sellers whose own photographs are stolen face the mirror-image problem, and our guide to copycat listings covers finding and reporting them.
An AI generated listing photos check takes about a minute once you have done it twice, and it is worth doing before every purchase from a seller you have not used before. The search has a bonus, too: it usually surfaces the genuine listing, so the technique that exposes the fake also shows you where to buy the real thing.
What is the generated-set drift check?
AI image generators are good at one image and bad at six. Ask a generator for the same product from six angles and it will change the product slightly each time, because it has no real object to keep consistent. A real item keeps the same shape, colour, pattern and number of parts from every angle, AnyFact’s marketplace guide states that drift test in exactly those terms, so comparing the photos in one listing against each other is the most reliable check in this guide. Here is how to run it.
- Open every photo in the listing side by side. Most marketplaces let you open the gallery full-screen; on a phone, save each photo and view them in your gallery.
- Pick three features to track. For most products, the handle or strap, the logo or label, and one distinctive detail, a zip, a clasp, a pattern, are enough.
- Follow each feature through every photo. Does the handle sit on the same side? Does the logo stay the same size and spelling? Does the pattern repeat identically, or does it change between shots?
- Check proportions against a fixed reference. If a mug appears beside the vase in one photo, the vase should be the same height relative to the mug in every photo where both appear.
- Count the drift. One small difference can be a photo angle. Two or three differences across the set, a handle that moves, a glaze that changes, a maker’s mark that appears and vanishes, is the signature of a generated set.
Worked example (illustrative). Take a typical marketplace listing for a “handmade ceramic vase”, six photos, no seller history. Photo one is the hero shot: a sage-green vase with a curved handle on the left, speckled glaze, on a wooden table. Photo two shows the vase from the other side, but the handle is now on the right and straighter, and the speckles have become a smooth gradient. Photo three is a close-up of the base: a maker’s mark stamped into the clay. Photo four shows the same base from a slightly different angle, and the mark has gone. Photo five places a mug beside the vase for scale; photo six shows the mug again, and the vase is now noticeably shorter relative to it. No single photo looks wrong on its own. Together, the set cannot be photographs of one object, because one object cannot change its handle, lose its maker’s mark and shrink between shots. That pattern, drift across the set, not a flaw in one image, is what the drift check for AI-generated product photos is built to find. This example is illustrative, assembled from the drift pattern AnyFact documents, and real listings vary: a genuine seller’s photos can differ in lighting and angle, so treat drift as a reason to ask for more evidence, not as proof on its own.
When the set drifts, ask the seller for one new photograph of the exact item, held in a hand, beside today’s newspaper, or next to an object you name. A seller with the real item can take that photo in a minute. A seller with only generated images cannot, and the silence is itself an answer.
Listing photos vs buyer photos: what should you compare?
Star ratings are easy to fake. Photographs taken by actual buyers are much harder, and most marketplaces let you filter reviews to show only the ones with images. Those snapshots, shot in someone’s kitchen under bad lighting, tell you what the product really looks like once the studio magic wears off, and they are the closest thing to a photograph of what will actually arrive.
| What to compare | Listing photos | Buyer photos |
|---|---|---|
| Colour | Rich, saturated, evenly lit | Often duller or a different shade entirely |
| Size and scale | Shot to look generous, rarely with a reference object | Shown next to hands, tables and everyday objects |
| Material and finish | Smooth, flawless, premium-looking | Reveals plastic, printed veneer or thin fabric |
| Shape and details | Consistent within the generated set only | Shows the real seams, zips and proportions |
| Background | Studio scene, lifestyle room or plain white | A real home, with ordinary clutter and lighting |
Line the buyer photos up against the listing’s official images, because that is where the gap shows. The rich navy in the listing turns out to be a dull grey. The “solid wood” looks like printed veneer. The portion size is half what the hero shot implied. A listing with a slick hero shot and zero real buyer photos is a quiet warning on its own, and a listing where the buyer photos show a different item from the one pictured is a listing to leave alone.
Food photos have the same problem in a different aisle: our AI food photos check covers menu and delivery app pictures, where the stakes are a meal rather than a purchase.
Be a little sceptical of review photos that themselves look like polished marketing shots, since fake reviews sometimes recycle stock images. Real buyer photos are imperfect, and that imperfection is exactly what makes them trustworthy. It is also worth remembering that the US Federal Trade Commission’s rule on fake reviews and testimonials, finalised in August 2024 and effective from 21 October 2024, bans fake reviews, including reviews generated by AI, which is one more reason to read the photos in reviews rather than the star count.
If the product photo does not match the item received, that gap is your evidence. Photograph what arrived, keep screenshots of the listing as it stood, and contact the seller first, then your card provider if the seller will not help. The next section sets out the order to do it in.
The photos don’t add up, what should you do next?
A failed photo check is not a reason to argue with the seller. It is a reason to stop, gather one more piece of evidence, and then either walk away or buy with protection. Work through these steps in order.
- Ask for a fresh photo of the exact item. Request one photograph you choose, the item held in a hand, beside an object you name, or with today’s date on a piece of paper beside it. A genuine seller can do this in a minute; a seller working from generated images cannot. No reply, or a reply with another polished studio shot, ends the enquiry.
- Run the clearest photo through a detector as a second opinion. AI image detectors estimate the probability that an image was edited or generated, and they are useful as a tiebreaker when your eyes cannot decide. Our guide to the best AI image detector tools compares the free options and explains how to read their verdicts, as a second opinion, never as proof.
- Check the shop, not just the photo, if it is a standalone website. This guide judges the photo on a platform listing; whether an unknown shop with its own website is legitimate is a separate check, and our guide to checking whether an online shop is real covers the business, domain and payment checks for that case. Do not mix the two: a real shop can use a bad photo, and a fake shop can steal a good one.
- Pay by card if you do buy. The US Federal Trade Commission advises that paying by credit card best protects you and your money in case of a scam, and warns never to buy from sellers who insist on gift cards, wire transfers or cryptocurrency, those are payment methods from which money typically cannot be recovered. The FTC’s online shopping guidance sets out the same advice alongside how to compare sellers and keep records of a purchase.
- Report a listing that fails the checks. Use the marketplace’s own report route, every major platform has one for misleading listings, and include the listing URL, the photos, and what your checks found. Reporting protects the next buyer, and platforms act faster on reports with evidence attached.
- Know the wider pattern. A generated product photo is one trick in a larger set of photo scams, from fake celebrity endorsements to stolen listing photos. Our complete guide to photo scams maps the whole landscape, so you can recognise the next variant before it costs you money.
If you have already paid and the item does not match the photo, act quickly: contact the seller, then your card provider about a chargeback, for UK credit card purchases over ยฃ100, Section 75 may make the card provider jointly liable, and keep every screenshot, email and receipt. The photo check takes a minute; the return takes weeks. Do the check first.
FAQs
Can a detector always tell if a product photo is AI-generated?
No, not always, and you should not treat it as a final verdict. A detector estimates the probability that an image was edited or generated, and the newest AI models can occasionally slip past one. Use it alongside a reverse image search and the buyer photos, and let the combined picture guide you rather than any single tool.
Is a reverse image search enough on its own?
It is the best single check, but it is not foolproof. Reverse search reliably catches stolen photos, stock images and shots duplicated across many sellers. What it can miss is a brand-new AI-generated image that has not been indexed anywhere yet, which is why you pair it with the visual tells and a forensic detector.
The product photo doesn’t match the item I received. What should I do?
Photograph what arrived, keep screenshots of the listing as it stood, and contact the seller first with both. If the seller will not help, contact your card provider about a chargeback, and for UK credit card purchases over ยฃ100 ask about Section 75. Report the listing to the marketplace with your evidence attached.
Is one AI-looking detail enough to call a listing fake?
No. Judge the pattern, not one detail: a single warped label could be compression, a bad crop or a genuine photo edited for brightness. Ask the seller for a fresh photo of the exact item, run the reverse search and the set comparison, and only walk away when two or more checks point the same way.







