By Alexander Jordan, Founder
Published: November 11, 2026
When someone asks ChatGPT for linen sheets, queen size, under $300, nothing browses your homepage. The agent filters. It keeps the products whose data says linen, queen and under $300, and drops the rest. If your product data doesn't say linen, you aren't in the answer, however good your sheets are.
So we checked what 20 Shopify brands actually tell a machine about their products. Not what the page says to a shopper. What the data says to an agent.
The sample: 20 Shopify stores, four in each of five categories: furniture, apparel and footwear, beauty, bedding and home textiles, and food and beverage.
The products: the first ten real products in each store.
The yardstick: OpenAI's product feed spec for ChatGPT (Source: OpenAI, Agentic Commerce product feed spec, checked 4 November 2026).
The bar: a detail counts as there when it appears on at least 8 of the 10 products, somewhere a machine can read it: a product field, a variant option, a tag or the structured data on the page. Descriptions are logged separately, because a sentence in a paragraph is something an agent has to interpret, not something it can filter on.
Twenty stores is directional, not a census. Everything comes from each store's public product data as of 1 November 2026.
Material only matters to a buyer in some categories, so we counted it in three of them: furniture, apparel and bedding. That's 12 brands.
Filterable, 6 of 12: material sits in a tag, option or field. When a buyer names a fabric, these brands make the cut.
Description only, 3 of 12: it's in a sentence, not a field. An agent has to read it and guess.
Barely there, 3 of 12: stated on fewer than 8 of their 10 products, anywhere at all. To an agent, a product with no material isn't made of anything.
Bedding is the surprise. Cotton or linen is the whole decision when you buy sheets, and none of the four bedding and home textile brands put material anywhere an agent can filter on it.
Furniture did best. All four furniture brands put material in tags or options.
The first question anyone asks about a sofa is whether it fits. None of the four furniture brands stated dimensions on most of their products, anywhere we could read, description included. One uses its "Size" option for configuration names, not measurements.
Dimensions are optional in OpenAI's spec. They're also exactly what a buyer types when they say "under 80 inches".
On Shopify, the vendor field says who makes the product. OpenAI's spec asks for the brand as shown on the product page.
Kylie Cosmetics: the vendor reads "kylie-cosmetics" and "kylie-fragrance", lowercase and hyphenated, not the brand name a buyer would type.
Death Wish Coffee: its own name appears two ways ("Death Wish Coffee Company" on coffee, "Death Wish Coffee" on a mug), and its coffee bundles name a pottery maker as the vendor.
When the data says one maker and the page says another, an agent has two answers and no way to know which is right.
Private fields: we only read what anyone outside the store can read: the public product data, tags, options, the structured data on the page and the description. If a brand keeps material or dimensions in fields that aren't public, we wouldn't see them.
Ingredients: we didn't check them for the beauty and food brands, so those categories aren't counted on material.
Shopify Plus: it can't be proven from outside the store, so we call these Shopify stores.
Truth: we checked whether a detail is there, not whether it's true. That's a separate job.
This isn't SEO. A material field won't help you rank for "linen sheets". It decides whether you're still in the running when the buyer says "linen".
Agents read your data, filter on it and move on. So the fix is plain. Say what the product is made of and how big it is, in a field, in plain words, taken from what your own product page already says. Keep marketing language out of the values.
In several cases the information is already on the page, in a description or an image. It just isn't anywhere a machine can filter on. A shopper reads your page. An agent reads your data.
For the full picture of what AI shopping agents read, start with our guide to agentic commerce optimization for Shopify Plus. Next Wednesday: the fields agents filter on, Shopify's standard category attributes, and how to fill them from your own page evidence.
Sample: 20 Shopify stores, 5 categories, 10 products each, public data as of 1 November 2026.
Measured against: OpenAI's product feed spec for ChatGPT, checked 4 November 2026.
Material: 6 of 12 furniture, apparel and bedding brands put it where an agent can filter. For bedding alone, 0 of 4.
Dimensions: 0 of 4 furniture brands state them on most products.
Vendor: 2 of 20 name the wrong maker or spell their own name two ways.
Limits: public data only, ingredients not checked, directional, not a census.
