Too Perfect to Be Real: A Buyer's Field Guide to AI-Faked Listings

You spot a mid-century anglepoise lamp on Facebook Marketplace for thirty-five quid. The photos are gorgeous: soft window light, a tasteful bit of exposed brick, not a speck of dust on the shade. The description is fluent, friendly and oddly free of detail. The seller joined three weeks ago.

None of that is proof of anything on its own. Put together, though, it is the shape of a listing that has been assembled rather than photographed. And that shape is getting more common.

Why the fakes got so convincing

A couple of years ago, AI-generated product shots gave themselves away instantly — six-fingered hands, melted brand logos, a kettle with two spouts. Those errors have largely been sanded down. What is left is subtler: images that are technically flawless and emotionally wrong.

A real second-hand listing looks like someone's hallway. There is a bit of carpet, a shadow from a coat, a reflection in a mirror of a person holding a phone. Synthetic images tend to look like a showroom that has never been lived in, because that is what they were trained on.

Reading the photos like a sceptic

Start with the small stuff. Zoom in on labels, packaging and stitching. Look at whether text on a box resolves into actual words or into plausible-looking shapes. Check reflections in glass and metal — do they match the room? Look at shadows: if the lamp casts a soft shadow but the table beneath it casts a hard one, something is off.

Then look at the set as a whole. Genuine sellers take four photos in the same room in the same five minutes. If the item appears against four completely different backgrounds, or the same background appears across several different sellers, treat it as a question worth asking.

The reverse image habit

This is the single most useful trick available and it takes about fifteen seconds. Save the photo, then run it through Google Lens or a similar reverse image search. If it comes back attached to an AliExpress listing, a US auction site, or three other UK sellers in different counties, you have your answer.

Do this for anything over about fifty pounds, and do it before you get emotionally attached to the item. Bargain hunting is 90 per cent dopamine.

Ask for something a bot cannot produce

The classic test is to ask for a photo of the item next to a piece of paper with today's date and your name on it. It still works, but it is no longer airtight — image generators can render handwriting reasonably well now.

So make it awkward. Ask for the item beside their kettle with the microwave clock visible. Ask for a ten-second video of them turning it over and reading out the serial number. Ask what the postcode district is and whether collection is possible this week.

Genuine sellers do this in under a minute, sometimes with a slightly grumpy message about time-wasters. Scammers stall, change the subject, or produce something that answers a slightly different question to the one you asked.

Reviews arrive in batches

Fake feedback rarely looks random, because randomness is hard to fake. Watch for a cluster of five-star reviews landing on the same day, or a fortnight of them followed by nothing for a year. Read the wording: real reviews mention specifics — a chipped corner, a fast dispatch, a confusing instruction manual. Synthetic ones say things like 'Great seller, thanks' in five slightly different ways.

On eBay, look at feedback left for others as well as feedback received, and check whether the account has been buying twenty-quid items for years and suddenly selling iPhones. On Vinted, an account with plenty of reviews on cheap clothing and a sudden burst of sealed electronics is worth a second look.

The money is where the scam happens

Here is the bit that actually matters. The fake photo is bait; the payment is the trap. Every protection you have depends on paying through the platform.

Bank transfer offers you nothing. PayPal friends and family offers you nothing by design — that is the whole point of it. A screenshot of a payment that has not cleared is not a payment. A message saying the courier needs a small fee before release is a second scam layered on the first.

On the buying side, a credit card gives you Section 75 protection on purchases over £100, and debit card payments can sometimes be charged back. On the selling side, never post anything until the money is actually sitting in your account, and be suspicious of buyers who are strangely relaxed about price.

If you think you have been caught out

Screenshot everything before you do anything else — the listing, the messages, the payment confirmation. Report the listing to the platform using its fraud route rather than a general complaint form. Report it to Action Fraud in England, Wales and Northern Ireland, or Police Scotland north of the border. Ring your bank the same day; the sooner a payment is flagged, the better the chance of recovering it.

If you posted an item to a buyer who then vanished, contact the courier and ask about interception. It rarely works, but it costs nothing to try.

A sixty-second check before you tap Buy

• Reverse image search the main photo
• Zoom in on labels, logos and reflections
• Check when the account was created and what it has sold before
• Read three reviews properly rather than skimming the star rating
• Ask for a fresh photo or video with something specific in frame
• Pay through the platform, every time, no exceptions
• If the price is half the going rate, ask yourself who is subsidising it

The internet has always rewarded a bit of healthy suspicion. AI has simply made that suspicion cheaper to satisfy — which, on balance, is good news for anyone prepared to use it.