It's a Sunday evening in November. A husband opens an AI assistant, uploads three photos and types: "Help me with the Christmas list. Here's what my kids and my wife asked for. Where can I find these, and where are they cheapest?"
Photo one is a seven-year-old's drawing of "a sparkly green dress." Photo two is a screenshot of gold hoop earrings. Photo three is a pair of sneakers torn out of a catalog, slightly crumpled, with a coffee ring on it.
He has done his job. He has delegated. He is, in his own words, "basically done with Christmas."
The assistant can see the photos. But when it goes looking in stores, it reads mostly words: the product title, the description, and the color and size in the product data. It searches for green, sequin, girls' size, gold, hoop.
Somewhere, two streets from his house, there is the perfect dress at the right price. Its product page says "Party dress." That's it. That's the whole page.
It doesn't make the list. He buys a different dress from a brand that wrote one more sentence. Christmas is saved. Your sale isn't.
AI shopping assistants like ChatGPT, Gemini and Perplexity find products by matching the shopper's words to the words on a product page. If the title, description and product data don't name the color, material, fit and occasion, the assistant has nothing to match, so the product doesn't get recommended, even when it is in stock and a perfect fit.
A beautiful photo doesn't save you here. A page that only says "Party dress" is invisible to a shopper asking for a sparkly green dress. The fix isn't a new website. It's a few precise words per product, written where search, filters and AI can read them.
Shoppers can't buy what they can't find. Neither can their AI.
Fashion brands spend most of their energy getting people to the store. The expensive part happens after they arrive: the shopper can't find the product, or finds it and guesses wrong.
Both problems have the same root. The product page is missing the words the shopper, the filter and the AI are looking for.
1. Shoppers can't find what's already in stock. A Google Cloud survey run by The Harris Poll put the cost of shoppers leaving after a failed search at more than $300 billion a year in the US alone. You paid for that visit. The shopper left with nothing.
2. Shoppers who do buy, guess. Apparel has the highest return rate in e-commerce, around 25% on average, with size and fit behind roughly half of returns. When the page doesn't say how a piece fits or what it's for, the shopper finds out in their hallway and sends it back.
3. Now AI is doing the shopping, and it reads even more literally. In the 2025 holiday season, traffic to US retail sites from AI tools rose 693.4% year over year, and those AI visitors converted 31% better than other traffic. The best-converting shoppers of the season arrived through a channel that only sees what's written.
The stakes peak right now. Cyber Monday 2025 was the biggest online shopping day of the year, with $14.25 billion spent online in the US. This year, Black Friday is 27 November and Cyber Monday is 30 November.
A photo can be gorgeous. AI still can't search a vibe.
Same dress, same photo, same price. Only the words change, and that's what decides whether it makes the Christmas list. Here is what the AI assistant sees on each version (an invented example).
| What the AI reads | Before | After |
|---|---|---|
| Title | Party dress | Girls' Sequin Party Dress, Emerald Green |
| Color | (none) | Emerald green |
| Material | (none) | Sequins on soft tulle, cotton lining |
| Size | S, M, L | Girls' ages 4–10 |
| Fit | (none) | Relaxed bodice, twirly knee-length skirt |
| Occasion | (none) | Christmas parties, holiday photos, weddings |
| Description | Perfect for every celebration. | Made for twirling. Emerald sequins catch the tree lights, and the cotton lining means no itchy complaints by dessert. |
The husband typed sparkly, green and dress. The "before" page matches one word out of three. The "after" page matches all of them, plus the size his daughter needs.
Notice what didn't change: the brand's tone. Good product data doesn't sound like a spreadsheet. It sounds like you, with the facts filled in.
Six details decide whether shoppers, filters and AI assistants can find a product. Each one should be written in the title, description or product data, not only shown in the photo.
Test it in five minutes: ask ChatGPT or Gemini for one of your products the way a gift shopper would, for example "a gold hoop earring under $80 for my wife." If your product doesn't come up, the assistant didn't find enough words to match.
Catecut is a Shopify app that scores your product pages, then fills in what's missing in your brand's own voice. It's built for fashion and jewelry brands with more products than hours in the day.
Doing the checklist by hand is fine for 10 products. By product 147 it's midnight and every dress is "lovely." Catecut does product 147 like product 1:
Nothing goes live until you've reviewed it. AI does the heavy lifting. You keep the steering wheel.
Strong product pages take a team. Catecut is yours.
AI assistants recommend products whose pages clearly match the shopper's words. If your title, description and product data don't name the color, material, fit and occasion, the assistant has too little to match and picks a page that does.
They can see photos a shopper uploads, but when they search stores they rely mostly on written text: titles, descriptions and product data. A beautiful photo with a one-word title is still hard to find.
Color tone, material or metal, fit, occasion, and a title that says what the product is. For jewelry, add the closure type and whether a piece is plated or solid.
It overlaps. The same details help Google, your store's own search and filters, and AI assistants. Fixing them once improves all four.
No. Many brands keep their descriptions and only fill in the missing details, Google titles, meta descriptions and alt text.
By hand, a few minutes per product. With Catecut, a whole catalog can be drafted in your voice and reviewed in days, not weeks.
Somewhere tonight, someone is asking an AI for a sparkly green dress, gold hoops or the perfect sweater for their dad. Make sure your product is one it can find.
Install Catecut and see your Listing Score for every product. It's free, no credit card needed.
Don't take our word for it. Take your score.