Since our coverage of Meta's expanding AI ambitions earlier this month — Ray-Ban facial recognition, an AI-generated news feed that got quietly killed, and a reported multi-billion dollar share sale — a ground-level story has been unfolding on the company's own platform that nobody in the boardroom planned for.

Ordinary sellers figured out how to game Facebook Marketplace with AI. And the results are hard to argue with.

How It Started

On May 28, journalist Trung Phan posted on X about a seller named Brandon Carone who was trying to move Caterpillar 950 forks and a dump bucket — combined asking price: over $10,000. Heavy industrial equipment. Not exactly a high-demand consumer category.

Carone's solution was to use AI image generation to superimpose female models in shorts and nightwear directly onto the product photos. The listings went viral on X, racking up nearly 3 million views, according to Dexerto.

Within hours, other sellers were copying the playbook.

A user going by Simon Paquin posted that he tried the same tactic with gym machinery. Result: "flooded with interest today." His follow-up: "The listings were dead until I updated the images."

Multiple sellers reported the same outcome.

This Isn't New Marketing — The AI Part Is

Using attractive imagery to sell mundane products is as old as advertising itself. Car shows have done it for decades. Tool calendars have done it for decades. The difference here is the cost and the barrier to entry.

Before AI image generation, pulling this off required hiring a model, a photographer, and a graphic designer. Now it takes five minutes and a free chatbot.

A seller of used Caterpillar parts in the middle of nowhere can now run the same visual playbook as a national brand. Like it or not, that's a real shift.

What Meta Is NOT Doing About It

Meta has spent enormous resources building content moderation systems. The company has repeatedly claimed it can detect and remove policy-violating content at scale. It banned thousands of accounts — many of them without due process, as we reported on June 4 — in the name of keeping the platform clean.

And yet AI-generated thirst-trap imagery sitting inside product listings for heavy excavator attachments remains up and going viral.

Meta's community standards prohibit nudity and sexual solicitation. But "scantily clad" AI models posing with dump buckets apparently doesn't cross whatever threshold their moderation tools are set to catch. Either the filters aren't looking for this, or they looked and decided it doesn't violate policy.

Neither scenario is reassuring for a company that just announced it wants its AI to be the backbone of how people consume news and shop online.

The Algorithm Is the Product

What this episode reveals is something fundamental about how Facebook Marketplace works.

Meta's algorithm rewards click-through rates. More clicks equal more visibility. The platform doesn't particularly care why someone clicked — only that they did. When an AI-generated image of a woman in shorts next to a $6,500 piece of excavator equipment generates more clicks than a plain photo of the equipment, the algorithm treats that as a quality signal and boosts the listing.

The platform is being gamed by its own incentive structure. And Meta built that structure.

This is the same company that, as we reported on June 6, built an AI-generated clickbait news feed inside its app and had to kill it after getting caught. The pattern is consistent: build systems optimized for engagement, watch engagement get gamed, react after the fact.

A Broader Problem

Most coverage of this story — from Dexerto and the X discourse — treated it as a funny viral moment. "Genius sellers!" "Hilarious!" And it is.

But this is a preview of what AI-assisted manipulation of algorithmically ranked platforms looks like at scale. Today it's a guy selling Caterpillar forks. Tomorrow it's someone using the same technique to drive traffic to a scam listing, a phishing link, or a political ad disguised as a product.

Meta's moderation team is already stretched. It already makes large-scale account banning errors. It already got caught pushing AI-generated news. Adding a new vector — AI-manipulated product imagery optimized for click-through — is a genuine problem.

What Comes Next

Facebook Marketplace sellers cracked an engagement hack that Meta's own systems can't stop. The hack works because Meta's algorithm doesn't reward honesty — it rewards clicks. And AI just made that hack available to anyone with a phone and five minutes.

Brandon Carone wanted to sell some excavator parts. He figured out something about Meta's platform that Meta's own engineers apparently haven't fixed. Whether that's funny or alarming depends on what you think these platforms are actually for.