By Steve Merrill · October 26, 2026
AI shopping assistants read your product feed before they recommend you. The seven fields that decide your fate are title, description, GTIN/MPN/brand, price and availability, product_type and google_product_category, image_link, and product_highlight.
I ran a $10M clothing brand for 15 years. I know how easy it is to treat the product feed like a chore and let it rot. That's the mistake. The feed is the thing AI actually reads. The model picks winners from structured feed data before it ever looks at your homepage. Here are the seven fields that matter.
1. title
The title is the first thing AI parses, and it carries the most weight of any field. Google's product data spec allows 150 characters, but the first 70 do the heavy lifting. Lead with brand, then product, then the attribute a shopper types. "Nike" is weak. "Nike Pegasus 41 Men's Running Shoe, Black, Size 10" is what gets matched. The pattern the model wants is brand, model, then the two attributes people search by.
2. description
AI uses the description to understand what the product does and who it's for, not just what it's called. Write it in plain sentences with the real use case and materials, because the model reads it the way a shopper would. Keyword stuffing here gets you ignored. Facts get you matched. A line like "waterproof to 10 meters, fits wrists 6 to 8 inches" gives the model something to answer a question with. "Premium quality" gives it nothing.
3. gtin / mpn / brand
The GTIN is the universal barcode ID that lets AI tie your product to the same item everywhere else on the web, including its reviews and specs. Google treats missing or wrong GTINs as a top reason to suppress a listing. If you make your own products, the brand plus MPN combo does the same job. Fill all three when you can. One wrong digit and the model matches you to a different product, with someone else's reviews.
4. price and availability
Price and availability are the two fields AI trusts least when they're stale, so it checks them hardest. If your feed says $49 and in stock but your Shopify product page says $59 and sold out, the model drops you to protect the shopper. Sync these in real time. A mismatch here kills recommendations faster than anything else on this list. With agentic checkout, the model expects the feed to match the cart at the moment of purchase.
5. product_type and google_product_category
These two fields tell AI exactly where your product sits so it surfaces you for the right question. The google_product_category uses Google's official taxonomy, and product_type is your own store's breadcrumb. Get the category wrong and you compete in the wrong bucket. Get it right and you show up when someone asks for exactly what you sell. File a running shoe under "Apparel" instead of "Athletic Shoes" and you never appear for the shopper who wants running shoes.
6. image_link
AI assistants now read images, so the main image_link does double duty as data and proof. A clean, high-resolution shot on a plain background reads better than a busy lifestyle photo with text slapped on it. Google rejects images with promotional overlays and wants at least 800 by 800 pixels. A rejected image means a suppressed product, and a sharp shot reads as a trust signal that you're a real store.
7. product_highlight and product_detail
These structured fields let you hand AI the specs in a format it can lift straight into an answer. product_highlight is your short bullet points, and product_detail is attribute-value pairs like "Material: Merino Wool." This maps directly to the schema.org Product properties AI already expects. Fill them and you stop making the model guess. Ask "is this machine washable," and the answer sits in a product_detail pair ready to quote instead of buried in a paragraph.
How We Chose This List
These seven fields come from Google's own product data spec and from watching which Shopify feeds actually get surfaced by AI shopping assistants versus which get ignored. We ranked them by how often a missing or wrong value causes a product to drop out entirely.
FAQ
Q: What product feed fields do AI shopping assistants care about most?
Title, description, GTIN/MPN/brand, price and availability, product_type and google_product_category, image_link, and product_highlight. These give AI the structured facts it needs to match you to a shopper's question.
Q: Does a GTIN really matter for AI recommendations?
Yes. The GTIN ties your item to the same product across the web, including its reviews and specs. Missing or wrong GTINs are a top reason products get suppressed in Google Merchant Center.
Q: How long should a product title be in the feed?
Google allows up to 150 characters, but the first 70 carry the most weight. Lead with brand, then product, then the attribute a shopper would actually type.
Q: What breaks a product feed fastest?
Stale price and availability. If the feed and your Shopify store disagree, AI drops you to protect the shopper. Keep it synced in real time.
Your feed is either feeding AI or starving it. There's no middle. If you want us to audit your Shopify product feed and fix the fields that are costing you recommendations, see how WRKNG Digital gets stores recommended by AI.
