Everyone can spot AI-written text now.

The flat phrasing. The three-part structure. The bullet points that explain what the paragraph already said. We've trained ourselves to notice it.

But show the same person an AI-generated image — styled with a long prompt, dramatic lighting, a Japanese model in editorial fashion — and suddenly it's "incredible." "How did they make that?"

The output didn't change. The audience's vocabulary did.

Text is a medium everyone uses daily. We read, we write, we know what sounds human. Visual design, photography, typography — most people don't have the language for it. They can feel "good" or "off," but they can't say why.

That gap is where visual AI slop hides.

And it's being monetized. Long prompts sold as secrets. Custom GPTs sold as strategy tools. The promise is not better output. It is confidence, packaged as method.

The placebo works because the user can't tell the difference. Not because the prompt was magic.

A 2,000-word prompt full of "forbidden" rules and "emotional design" instructions looks impressive. It also mostly tells the AI things it already knows, or things vague enough that no one could verify whether they worked.

"Create tension between the photo and the text." Sure. How would you know if it did?

The prompts are written for the buyer, not the model.

This is not unique to images. "Websites that convert." "Copy that sells." The same dynamic applies. Show someone a site built with AI and a long template prompt, and they'll say it looks professional. Show them the same quality of writing in a Word doc, and they'll say it feels generic.

Medium changes perception. Vocabulary — or the lack of it — determines what we call slop.

What is actually being sold in most "AI monetization" products is not a tool. It is a reduction in friction. The buyer does not yet feel comfortable asking AI directly. The seller packages that ask, wraps it in a framework, and charges for the confidence to begin.

The irony of "AI monetization" is that many of the products being sold are things AI can already generate for the buyer.

What is being sold is not expertise. It is a shortcut to the feeling of expertise.

The real question isn't whether the prompt worked.

It's whether you could tell if it didn't.