By Steve Merrill, Founder of WRKNG Digital | August 15, 2026
Which product feed fields decide whether AI recommends you?
The fields that decide it are title, GTIN, brand, google_product_category, description, price, availability, condition, image_link, and review ratings. Get those ten clean and complete and AI shopping assistants can match, trust, and surface your product. Miss them and you're invisible, no matter how good the product is.
AI doesn't guess. It reads structured data. Here's the exact ten, why each one matters, what good looks like, and the mistake that kills you.
1. Product Title (title)
The title is the single most important match signal in your feed. AI reads it first to decide if your product answers the shopper's question. Good looks like Brand + Product + Key Attributes ("Bella Ella Ribbed Cotton Midi Dress, Olive, Women's M"), not "Best Seller Dress." The Google Merchant Center title spec caps it at 150 characters. The common mistake is stuffing marketing words instead of the attributes AI actually matches on.
2. GTIN and MPN
The GTIN (barcode number) is how AI knows your product is the same one sold elsewhere, so it can pull reviews and compare prices. Branded products need a valid GTIN. Manufactured products without one need the MPN plus brand. Per Google's unique product identifier rules, missing or wrong GTINs get products filtered out. The common mistake is faking a GTIN or leaving it blank when your product has one.
3. Brand
Brand ties your product to an identity AI can recognize and rank. It's required for almost every category with branded goods. Good looks like the actual manufacturer or your own brand name, spelled consistently everywhere. The common mistake is putting your store name in the brand field when you resell someone else's product, which breaks the match.
4. Google Product Category and Product Type
google_product_category places your product in Google's official taxonomy so AI understands what it is. product_type is your own breadcrumb path for extra context. Good looks like a specific leaf category ("Apparel & Accessories > Clothing > Dresses"), not a top-level guess. The product category field spec lists every valid value. The common mistake is picking a broad category or skipping it, which buries you against precise competitors.
5. Description
The description is where AI extracts the facts that answer specific shopper questions: fabric, fit, sizing, materials, use case. Good looks like plain, factual sentences packed with real attributes, front-loaded with the important ones. Google allows up to 5,000 characters. The common mistake is copying vague brand fluff or duplicating the title, which gives AI nothing new to match on.
6. Price and Sale Price
Price is a hard filter and a trust signal. AI assistants compare it directly and drop products where the feed price doesn't match the landing page. Good looks like the exact price a shopper pays, including currency, updated at least daily, with sale_price used for real markdowns. The common mistake is a mismatch between feed and site, which gets your product disapproved and demoted.
7. Availability
Availability tells AI whether it can actually recommend you right now. "in_stock," "out_of_stock," and "preorder" are the values that matter. Good looks like real-time stock status synced from Shopify inventory. The common mistake is stale availability. Recommend an out-of-stock product once and the assistant learns to stop trusting your feed.
8. Condition
Condition ("new," "refurbished," "used") sets shopper expectations and filters your product into the right query. Most stores are "new," and it's still required. Good looks like the accurate value set explicitly, not assumed. The common mistake is leaving it blank on refurbished or used inventory, which either hides you or gets you flagged for misrepresentation.
9. Image Link (image_link)
The image is how AI verifies your product is real and shows it to the shopper. Multimodal assistants now read the image itself, not just the URL. Good looks like a high-resolution main image on a clean background, no promotional text or watermarks, per Google's image link requirements. The common mistake is a low-res photo with a "SALE" banner burned in, which gets the image rejected.
10. Reviews and Ratings
Ratings are the strongest social proof AI uses to rank one matched product over another. Aggregate rating and review count push you up when two products otherwise tie. Good looks like real review data exposed through Schema.org AggregateRating on the product page and a connected product ratings feed. The common mistake is hiding reviews in an app widget AI can't read, so all that social proof is invisible to the assistant.
How we chose this list
These ten map directly to the required and recommended attributes in the Google Merchant Center product data spec, the fields Shopify pushes through its Google & YouTube channel, and the Product and Offer properties in Schema.org. They're the fields AI both requires to match a product and weighs to rank it. Nail these before anything else.
FAQ
Which product feed fields matter most for AI shopping recommendations?
Title, GTIN, brand, google_product_category, and price. They anchor your product to a real, matchable item with a clear identity and cost. Get those five clean before touching the rest.
Do I need a GTIN for AI to recommend my product?
For branded and manufactured products, yes. The GTIN lets AI match your listing to the same product across the web and pull in reviews and price comparisons. Custom or handmade products without one should set identifier_exists to false.
Does structured data on my Shopify product page replace the feed?
No. They work together. The Merchant Center feed carries product data to shopping surfaces. Schema.org Product and Offer markup on the page helps AI crawlers read the same facts straight from your site.
How often should product feed data update for AI?
Price and availability at least daily, near real time if you can. AI assistants drop or deprioritize products with stale or mismatched price and stock data.
Want us to audit your product feed and structured data so AI actually recommends your store? See how at WRKNG Digital's agentic commerce page.
By Steve Merrill, Founder of WRKNG Digital. August 15, 2026.
