By Steve Merrill, Founder of WRKNG Digital | August 5, 2026
AI recommends your store when your product feed is complete, accurate, and machine-readable. The nine fields that decide it are GTIN, brand, price, availability, condition, product_type, description, image_link, and review rating. Miss those and the assistant skips you, even when your product is better.
I've had a hard time getting people to care about feed data. It's boring. But AI shopping assistants can't see your beautiful product page. They read fields. Here's the nine that matter, in order.
1. GTIN
The GTIN is the global barcode number that ties your product to a single identity across the whole internet. AI uses it to match your listing against every other store selling the same item, then compares price, stock, and reviews. No GTIN means the assistant can't confirm what you're selling, so it plays it safe and recommends the store it can verify. Google's Merchant Center GTIN spec spells out the exact format.
2. Brand
The brand field tells AI who makes the product, which is how assistants answer queries like "best running shoes from Brooks." Leave it blank and you drop out of every brand-specific question a shopper asks. It maps directly to the Schema.org brand property, so fill it in your feed and your Product schema.
3. Price
Price is the field AI checks before it recommends anything, because most shopping prompts include a budget. The number in your feed has to match the number on your page, or the assistant flags a mismatch and moves on. Put it inside an Offer object with the currency, and keep it current.
4. Availability
Availability tells AI whether the item is in stock right now, and assistants refuse to recommend products they think are sold out. A stale value is the quiet killer here. I've seen stores lose spots on in-stock items because the feed said out of stock from a sync that ran three days ago. Update it in real time and use the standard InStock value.
5. Condition
Condition marks the product as new, refurbished, or used, and AI treats these as different products entirely. A shopper asking for "new" gets a different answer than one asking for "refurbished." Most Shopify stores sell new goods and skip the field, which is a mistake, because a missing condition reads as uncertainty. Set it explicitly to new.
6. Product Type
Product_type is your own category path, like Apparel > Men > Jackets, and it's how AI understands where your item sits in a catalog. It's different from Google's product category, and you want both. A clear type helps the assistant surface you for the right query instead of burying you under a vague label. Shopify's product type docs cover how to set it.
7. Description
The description is the plain-language text AI reads to understand what your product actually does. Assistants pull specifics from it, so materials, size, and use case beat marketing fluff every time. Write it for a person answering a question, not for a slogan. Keyword stuffing gets you nothing here.
8. Image Link
Image_link is the URL of your main product photo, and AI shopping surfaces are visual, so a broken or missing image drops you from the card entirely. The link has to resolve, the image has to be the product on a clean background, and it has to match the description. One dead image URL and you're invisible on a results screen built around pictures.
9. Review / Rating
Rating data is the tiebreaker AI uses when two products are otherwise equal. AggregateRating with a star score and review count gives the assistant a reason to pick you. With none of it, you're a blank card next to a competitor showing 4.8 stars and 900 reviews. Mark it up with Schema.org AggregateRating and keep the numbers honest.
How We Chose This List
These nine come from the fields AI shopping assistants read most: the required and recommended attributes in Google's Merchant Center product spec, plus the Schema.org Product and Offer properties that on-page crawlers pull. We ranked them by how often a missing or wrong value drops a store from a recommendation.
FAQ
Q: Which product feed field matters most for AI recommendations?
GTIN. It ties your product to a global identity AI can match against every other store selling the same item, so it can compare your price, stock, and reviews with confidence.
Q: Do I need Schema.org markup if I already submit a Google Merchant Center feed?
Yes. The Merchant Center feed feeds Google's systems, but on-page Product and Offer schema is what most AI assistants read when they crawl your product pages directly. Run both.
Q: How often should product availability update for AI assistants?
In real time or close to it. AI assistants drop products they think are out of stock, so a stale availability value costs you recommendations even when the item is sitting in your warehouse.
Q: Does adding review and rating data actually change AI recommendations?
Yes. AggregateRating gives an assistant a reason to pick you over an identical product. With no rating data, you're a blank card next to a competitor with 4.8 stars and 900 reviews.
Want your product data built so AI assistants recommend your store? That's what we do. See how at WRKNG Digital's agentic commerce page.
