All articles

Your catalog says every product is for everyone. The agent can't pick from that.

Shoppers describe themselves to AI assistants by what went wrong last time: the foundation that turned orange, the jeans that gapped. An agent narrows by ruling products out, and it needs evidence to do it. A catalog that never says who a product is not for gives it nothing to work with.

A black and white photo of a woman in a chair reading a fashion magazine, beside the text: Strong brands are clear. Clear brands are easier to understand. And brands that are easier to understand are easier to recommend.

Pick ten products from almost any fashion or beauty catalog and read the descriptions back to back.

Flattering on every body. Suits all skin tones. Works for day or night. Safe for all skin types.

Every product is for everyone. Nothing is ever wrong for anybody.

I understand why. No merchandiser wants to write the sentence that talks a customer out of a sale. But that habit was built for a person browsing a page. An agent reads it very differently, and I think it is quietly costing brands recommendations.

She describes herself by what went wrong

Listen to how people actually talk to a shopping assistant once they get past the first vague request.

“Every foundation I buy turns orange on me by lunch.”

“Jeans always gap at the back of my waist.”

“I can’t wear anything scented near my eyes.”

“Blazers swallow me. I have a short torso.”

None of those name a product feature she wants. Each one is a fact about her, told as a failure. This is the person half of the query, and a lot of it arrives as a no.

Most people can’t tell you their undertone. They can tell you exactly which shade made them look ill.

Picking is mostly ruling out

An agent starts with far more candidates than it can show and has to get down to two or three it can defend. Most of that work is cutting.

Her “no” statements are what it cuts with. And a cut needs evidence the same way a pick does.

So take the shopper whose foundation turns orange. She has medium olive skin and gets oily through the center of her face. Here are two listings the agent might be looking at.

Listing one: “A weightless foundation that flatters all skin tones. 40 shades. Natural finish. $42.”

Listing two: “Shade 230 is a medium with a neutral olive undertone. It does not pull warm as it wears. Satin finish. On very oily skin it needs powder by midday. Not a good match for dry or flaky skin, where it tends to cling. $42.”

“Flatters all skin tones” can’t be tested against the one thing she told it. The agent can’t rule the product out, but it can’t stand behind it either. It ends up as filler in the list, or dropped.

The second listing answers her directly. It also admits a limit, and that admission is doing real work. A listing that says where a product falls short is easier to believe about where it doesn’t.

That is the difference between indexed and chosen. Both foundations were found. One of them could be matched to a person.

The sale you think you are protecting

The fear is obvious. Write “not for dry skin” and you lose the dry skin customer.

You were going to lose her anyway. She would have bought it, watched it cling to her cheeks, and sent it back. Then she would have told her assistant, which remembers.

Meanwhile listing two stays out of her answer and becomes the clear pick for the olive, oily shopper, with a reason attached. One avoided return, one earned recommendation.

“For everyone” gets you neither. It keeps you in the catalog and out of the answer.

You already own this data

Most brands already have a detailed record of who each product is not for. It is sitting in the returns system.

“Too long.” “Gaped at the bust.” “Darker than expected.” “Broke me out.”

Every one of those is a mismatch between a product and a person, written down by the person. Read together, they describe the boundary of who the product suits. Almost no one feeds that back into the catalog.

A match has to be able to say no

This is where a model of the customer earns its place. Fit, undertone, and occasion are relational. A dress is not “too long.” It is too long on someone 5’2". A shade is not “orange.” It is orange on a cool olive complexion.

You can’t write those facts from the product alone. You need the product read in detail and a structured picture of the shopper to set it against.

That is the layer we build at Veristyle. Computer vision reads each product for more than 200 style, fit, and color traits. A model of the customer holds the other side: physical traits, skin tone and undertone, body shape, style, occasion, intent, geography. Matching the two tells you who a product is right for. It also tells you where the match stops, which is the piece most catalogs have never written down.

Other GEO work makes the product easier to read. This makes it possible to choose.

Try it on your most returned product

  1. Pull the top three return reasons for the product that comes back most.
  2. Turn each reason into a person. “Too long” becomes a height. “Too orange” becomes an undertone. “Too tight in the arms” becomes a build.
  3. Read your listing as that person’s agent. Is there one sentence that would have warned it off?
  4. Then flip it. Who does this product suit best, stated just as plainly? Is that written anywhere?

If the answer to both is no, the agent knows less about your product than your returns team does.

See what agents can and can’t tell about your products

A free AI visibility audit shows how assistants describe your products today and which shopper questions your catalog leaves unanswered. If you’d like to see the matching run on your own products, book a demo.

Related reading

See Veristyle on your catalog

Book a demoGet it on Shopify