On this page
Why AI product photos are a problem
An AI image of a product isn't always a scam. Brands make lifestyle shots and mock-ups with AI, and many say so. The problem is a generated picture standing in for the thing you will actually receive:
- Products that don't exist: a dress, a cake topper or a garden ornament that was never made, sold from a picture.
- Products that look nothing like the photo: a textured, embroidered or hand-made item in the picture, and a flat print or a cheap copy in the box.
- Second-hand items the seller doesn't have: a phone, a bag or a car shown in a generated photo to take a deposit.
- Hidden flaws: a real item photographed, then cleaned up with AI to remove scratches, stains or damage.
The last one is an edit rather than a generated image, and it needs different checks; more on that below. For the first three, the picture is the whole product, so checking the picture matters.
Checks for buyers
Before you pay, work through the listing as a whole. Each step takes a minute or two:
Compare the first photo with the rest
The first image is the one most likely to be generated. The real product often appears further down the gallery, in plainer photos. If it looks different there, believe the plain ones.
Find buyers' photos
Reviews with photos, questions and answers, and posts from people who bought the item show what actually arrives. A popular item with no buyer photos at all is a warning sign.
Read the specifications against the picture
Materials, sizes and techniques should match what you see. Embroidery that the description calls a print, or a "solid wood" table with no visible grain, means the photo and the product disagree.
Check the seller
How long have they been selling, what else do they sell, and do their other listings share the same too-perfect look? A new shop with hundreds of unrelated items is a pattern.
Search the image
A reverse image search can show the same picture in other shops, under other names, or as a stock image.
Weigh the price
A price far below what the item could cost to make is a reason to slow down, whatever the photos look like.
What the photos themselves can tell you
Product photos are staged, lit and edited even when they are real, so the usual signs need care:
Still worth checking
- Consistency: the same product should look the same in every photo, down to the stitching, the logo position and the number of buttons.
- Physical sense: straps that attach to something, handles that a hand could hold, a chair whose legs all reach the floor.
- Labels and packaging: text on tags, boxes and screens that matches the brand and the item.
- A look that fits the seller: a small hand-made shop with magazine-perfect studio photos of every item deserves a second look.
No longer reliable
- Mangled text and hands: modern generators get both right most of the time.
- A glossy, polished look: professional product photography looks like that too.
- Plain white backgrounds: real studios and generators both produce them.
- Missing photo details: marketplaces usually strip them from every upload.
Car and electronics listings deserve the same care. Look for door handles, badges, ports and buttons in the wrong place, and for a background that doesn't fit where the seller says they are.
When a real photo has been touched up
Not every misleading listing photo is generated. Sellers use AI tools to swap a background, remove a scratch, smooth a crease or make colours brighter. The product is real, but the photo hides how it really looks.
That is an edit, and it sits outside what a detector like ours is built to find: a real product photo with an AI background may come back likely real or inconclusive. For second-hand items, ask for a new photo taken now from another angle, with something in the frame that shows it is current, such as a note with your name and today's date.
Our guide to AI-edited photos covers what to look for around an edited area, and how to find earlier versions of a photo.
How Perspicax helps
Perspicax combines proven computer vision techniques with TrueEye 2.3, our state-of-the-art detector. It examines every region of a photo, down to the pixels, and answers in one of three bands:
Three possible answers
It is built and measured on fully generated, photo-like images. In testing, it caught 98.6% of images from the newest OpenAI, Google and ByteDance models on average, and wrongly flagged fewer than 1 in 2,000 original camera photos.
Measured the way Perspicax answers: an inconclusive answer counts as a miss. See every number
As a buyer, save the listing photo and drop it into the check on this page. As a seller, you can check images from a supplier before you list them, so your shop doesn't end up selling a picture.
For marketplaces and trust teams
At marketplace scale, nobody can look at every upload. A detector can, and it fits into the review flows you already run:
- Screen every upload through the Perspicax API and store the answer with the listing.
- Route, don't block: let likely real listings through, send likely AI-generated ones to review, and ask for another photo when the answer is inconclusive.
- Write a clear rule on AI imagery: lifestyle mock-ups and illustrations may be fine when disclosed, but the main photo must show the actual item.
- Ask for a real photo of the item in high-risk categories, from new sellers, and on listings a buyer has disputed.
- Verify sellers: identity, payment details and history matter more than any single photo.
- Watch for patterns: many listings from one account with generated photos, or the same image across several shops.
Keep false alarms in mind. In testing, fewer than 1 in 2,000 original camera photos were wrongly flagged, but no detector is perfect, so give sellers a way to respond with a new photo before you act. To talk through a set-up, contact us.
If you sell online: use AI without misleading buyers
AI can make a shop look better without making it dishonest. The line is simple: the buyer should receive what the main photos show.
- Show the actual item in the main photo, taken by you, with its flaws.
- Say so when a lifestyle image or a mock-up was made with AI, in the image or in the description.
- Don't use AI to remove damage, wear or stains from a second-hand item: it turns a fair sale into a dispute.
- Keep your original photos, so you can show what the item looked like if a buyer or the platform asks.
- Check supplier and dropshipping images before you list them: a generated supplier photo is a promise you can't keep.
Buyers who receive what the photos showed leave good reviews. Buyers who don't open disputes, and misleading photos break most marketplaces' listing rules, whether AI made them or not.
Check an image now
Drop, paste or link a photo. Perspicax checks every region of it with computer vision and TrueEye 2.3, and answers likely AI-generated, likely real or inconclusive.
Questions people ask
How can I tell if a product photo is AI-generated?
Compare it with the other photos in the listing and with buyers' photos, read the specifications against what it shows, check the seller and run a reverse image search. Then check the photo with a detector built for AI images.
Are AI product photos allowed?
It depends on the platform and the use. Many allow AI lifestyle images or mock-ups when they are disclosed, but the photos of the item a buyer will receive should show that item. Check the platform's rules.
Why does the item I received look different from the photo?
The listing photo may have been generated or heavily edited. Keep the listing, photograph what arrived, and use the platform's dispute or return process.
Can Perspicax check a product photo with an AI background?
It can check it, but a real product photo with only the background changed is an edit, so it may come back likely real or inconclusive. Perspicax is built for fully generated images.
Can a marketplace check every listing automatically?
Yes. The Perspicax API answers for each uploaded photo, so you can route listings in your own review flow: let likely real ones through and review the rest.
Is a white background a sign of AI?
No. Studios and sellers use plain backgrounds all the time, and generators copy them. On its own it tells you nothing.