The Fields Agents Filter On: How to Fill Shopify’s Category Attributes from Your Own Product Pages
16 min read

By Alexander Jordan, Founder
Published: November 19, 2026
Shopify already gives every product category a set of standard fields for things like material, size and color, and AI shopping agents filter on them. Here's what belongs in them, and the one rule I'd hold every value to.
When someone asks ChatGPT for a short-sleeve linen shirt in a medium, the agent has to get from thousands of shirts down to a handful, and Shopify says agents can use standardized product data as filters to do it. A shirt whose data says linen, short sleeve and medium stays in the running. A shirt whose data says nothing about fabric can be filtered out, however good the shirt is.
Last week we published the Product Data Gap Report, where we took 20 Shopify stores across five categories and checked what their product data actually tells a machine, measured against the product feed spec OpenAI publishes for ChatGPT shopping. The bedding result stuck with me. None of the four bedding brands put the material anywhere an agent could filter on it, and cotton or linen is the whole decision when you buy sheets. None of the four furniture brands stated dimensions on most of their products either.
So this week I want to get practical, and show you where those facts belong on Shopify and how to fill them in without guessing.
Every product on Shopify can be placed in one category from Shopify's Standard Product Taxonomy, which is Shopify's shared way of classifying products. Once you choose the category, Shopify attaches the attributes that belong to it, and calls them category metafields. For a shirt, that's size, neckline, sleeve length type, fabric, target gender and color, and a sofa and a set of sheets each get their own set. Your storefront filters have been able to use these fields since September 2024, so for most stores this isn't new. What's new is who's reading them.
Shopify says Shopify Catalog, its structured source of product information from Shopify merchants, is what helps your products show up accurately in AI shopping answers like ChatGPT and Microsoft Copilot. Catalog maps products to the same taxonomy, and in Shopify's words, once the data is standardized it "can also be leveraged by the agents as filters." So what those fields say decides whether your product is still in the running when the buyer gets specific.
When you choose a category, Shopify says "based on AI predictions, some category metafields are added by default." Catalog goes a step further, and Shopify says it "will infer missing attributes when the product data is clear enough." That's a reasonable place to start, but it's still a prediction, and a wrong prediction does more damage in a filter than anywhere else, because it puts your product in front of a buyer who asked for something else. Shopify can fill the field for you, but making sure the value is true is still your job.
We fill a field only when the product's own page supports the value, never from what that kind of product usually is. In practice it looks like this:
Material: if the page says "100% European linen", the value is linen. If the page only says "soft, breathable bedding", there's nothing to fill yet, so fix the page first and then the field.
Color: pick the standard color closest to the one the page names. If the photos show a color the copy never names, name it on the page first.
Size: use the size the variant actually sells, in the same words your size chart uses, so the field and the page never disagree.
Who it's for: only fill in target gender or age group where the page says who the product is for.
The wording: keep values literal. Linen, not luxuriously soft linen. Marketing language belongs in the description, not in a field an agent filters on.
Everything else: if the page says nothing, leave the field empty and add the fact to the page. An empty field costs you one filter. A wrong one costs you the buyer.
That's exactly the furniture gap from last week's report, and I wouldn't wait for the standard fields to cover it, because Shopify lists unit of measurement and dimension attributes among the current limits of its standard taxonomy, saying they "aren't currently available." Put the measurements on the product page in plain numbers with units, and then send them in the dimensions field of your product feed, because OpenAI's spec has one for length, width, height and unit. That's what a buyer means when they say "under 80 inches".
Category: put every product in the most specific category that fits. Shopify suggests one, and it's worth accepting only when it's right.
Attributes: fill in every attribute the product page supports.
Predicted values: check each value Shopify added against the page, and remove anything the page doesn't back up.
Variants: connect size and color to your variant options, so each variant carries its own value.
Dimensions: put them on the page with units, and in the dimensions field your feed sends.
Gaps: wherever buyers filter on something your page doesn't say, fix the page first, then the field.
The fields already exist, and the agents already read them. Most stores have either left them empty or let a model guess. Here's a quick check you can do today. Open one best seller in Shopify, read every attribute, and find the sentence on the product page that backs up each value. Any value without a sentence behind it is a guess, and an agent will treat it as a fact.
If you want the full picture of what AI shopping agents read on a Shopify store, start with our guide to agentic commerce optimization for Shopify Plus. The Product Data Gap Report has the full findings from the 20 stores, category by category. Next Wednesday I'll be writing about agentic checkout on Shopify, which lets a buyer pay with Shop Pay inside Muse, Meta's AI agent that can shop on their behalf, and about what that changes for your product pages.
What are Shopify category metafields?
They're the standard attributes Shopify attaches to each category in its Standard Product Taxonomy, like size, fabric and color for a shirt. They come with whatever category you choose for the product.
Do AI shopping agents use them?
Shopify says Shopify Catalog, which maps products to the same taxonomy, is what helps products show up in AI shopping answers like ChatGPT and Copilot, and that agents can use the standardized data as filters. So the values in those fields help decide whether a product makes the shortlist.
Should I let Shopify fill them in automatically?
Use Shopify's predictions as a starting point, then keep only the values your product page backs up. A wrong value can put a product in front of a buyer who asked for something else.
Where do product dimensions go?
Don't count on Shopify's standard attributes for them yet, because Shopify lists dimension attributes as not currently available. Put them on the product page in numbers with units, and in the dimensions field your product feed sends.
Alexander Jordan
Founder, Firon Marketing
15+ years scaling brands. A strategic partner for high-growth founders, focused on sustainable revenue models and long-term equity value.