By Steve Merrill, Founder of WRKNG Digital | September 20, 2026
To fix a product feed for AI shopping, correct it field by field against the Google Merchant Center product data specification: accurate titles, factual descriptions, valid GTINs, correct product types, live availability, matching price and currency, and clean images. Get those right and AI tools can read, trust, and recommend your store.
This is the fix list. We already named the problems in 11 product feed mistakes that keep AI from recommending your store. Now here's the correction for each one, with a quick before and after so you know exactly what "fixed" looks like.
1. How should I write the title field?
Put the brand, product name, and one key attribute in the first 70 characters, because AI tools read the title first to match a shopper's request. The Google Merchant Center title spec allows up to 150 characters, but the front of the string does the work. Before: "Blue Shirt." After: "Patagonia Better Sweater Fleece Jacket, Men's, Navy."
2. What belongs in the description field?
Write plain, factual sentences about material, fit, and use. AI shopping tools pull descriptions to answer specific questions, so marketing fluff gets ignored while facts get cited. Before: "You'll love this cozy must-have piece!" After: "Recycled polyester fleece jacket. Full-zip, stand-up collar, two zip pockets. Machine washable."
3. Do I need a GTIN, and how do I set it?
Add the real GTIN if one exists, because Google uses it to match your product against other sellers. A missing or fake barcode drops you out of comparison results entirely. Before: GTIN blank. After: GTIN "0193249083621" filled, or identifier_exists set to false when the product truly has no barcode, per the Merchant Center unique identifiers guide.
4. How do I fix the product type and category?
Map each product to a Google product category from the official taxonomy, and use your own product_type for internal grouping. The right category tells AI where your item lives in a catalog of millions. Before: google_product_category empty. After: "Apparel & Accessories > Clothing > Outerwear > Coats & Jackets," which matches the Google product category spec.
5. Why does availability keep breaking?
Set availability to reflect live inventory and let Shopify sync it automatically. Nothing kills trust faster than an AI recommending a product that's sold out. Before: availability "in stock" on an item you ran out of last week. After: availability "out_of_stock" pushed in real time through the Shopify Google & YouTube channel.
6. How should price and currency be formatted?
Write price as a number plus the ISO currency code, and make it match your landing page to the cent. A price mismatch between feed and page gets your product disapproved. Before: "19.99." After: "19.99 USD," identical to the price shown at checkout, following the Merchant Center price attribute spec.
7. What makes an image_link work?
Use a high-resolution image on a plain white background, at least 800 by 800 pixels, with no promotional text or watermarks. AI shopping surfaces reject busy or low-res images. Before: a 300-pixel lifestyle shot with a "SALE" banner. After: a clean 1200-pixel product photo, per Google's image requirements.
8. Should I fill brand and MPN for every SKU?
Yes. Fill the brand field for every product and add the manufacturer part number when you have it, because these confirm the item is real and matchable. Before: brand blank on half your catalog. After: brand "Patagonia," mpn "25528" on every SKU. I've watched stores recover from invisible to recommended just by fixing this one gap.
9. How do I handle product variants?
Group every color and size of a product under one item_group_id, then give each variant its own unique id. This tells AI the six items are one product in six flavors, not six unrelated listings. Before: six standalone rows. After: six rows sharing item_group_id "BETTER-SWEATER-M," each with its own id, as the variant attributes spec requires.
10. Which shipping and return fields matter?
Add shipping cost, delivery time, and a return policy, because AI shoppers weigh these before recommending. A store with no shipping data looks riskier than one that spells it out. Before: shipping and return fields empty. After: shipping "US:::5.00 USD" with a stated 30-day return window, set through Merchant Center account-level policies.
11. How do I keep on-page schema in sync with the feed?
Match your Schema.org Product markup to the feed so price, availability, and identifiers are identical on both. When they disagree, AI tools flag the conflict and trust neither. Before: page schema shows "29.99, InStock" while the feed says "19.99, out_of_stock." After: both read "19.99 USD, out_of_stock." Read that again if you skipped it. The mismatch is what's costing you.
How we chose this list
Every fix maps to a required or recommended field in the Google Merchant Center product data specification, cross-checked against Shopify's feed sync and Schema.org Product markup. We ordered them by how often we see each one blocking real Shopify stores from getting recommended.
FAQ
How do I fix my product feed for AI shopping?
Fix it field by field against the Google Merchant Center product data specification: accurate titles, factual descriptions, valid GTINs, correct product types, live availability, matching price and currency, and complete brand and MPN values. Then match your on-page Product schema to the feed.
Which product feed field matters most for AI recommendations?
The title. AI shopping tools read it first to match a shopper's request, so put the brand, product name, and one key attribute in the first 70 characters.
Do I need a GTIN for AI shopping?
Yes, if a GTIN exists for your product. Google matches products across sellers using the GTIN, and a missing or wrong one hides you from comparison results. If your product has no barcode, set identifier_exists to false instead of inventing one.
Does on-page schema need to match my product feed?
Yes. If your Product schema shows a different price or availability than the feed, AI tools flag the mismatch and stop trusting both. Keep price, availability, and identifiers identical across the feed and the page.
Want your whole feed audited and fixed so AI actually recommends your store? See how we do it at WRKNG Digital's agentic commerce program.
