By Steve Merrill, Founder of WRKNG Digital | June 30, 2026
Nine product feed fields control whether AI shopping assistants recommend your products. Most Shopify merchants have at least four of them wrong - not broken, just wrong enough to get filtered out before a recommendation ever surfaces. We audited over 2,400 products across Shopify stores. Only 11% had the structured data needed to be recommended by ChatGPT Shopping.
Here's exactly what the other 89% are missing.
1. Product Title
Your product title is the primary signal AI assistants use to match your product to a buyer's query. ChatGPT Shopping's NLP matching is built on title parsing. "Black Sneakers" tells it almost nothing. "Nike Men's Air Max 270, Black, Sizes 7-14" tells it the brand, product line, who it's for, colorway, and size availability - all in one field.
The formula: brand + product name + key attribute (material, color, or gender) + size range. Every word is matchable signal. Missing any of them means missed queries.
2. Product Type (Google Taxonomy)
Google publishes an official product taxonomy with over 8,000 categories. That taxonomy is the standard AI shopping platforms use to classify products. "Apparel & Accessories > Clothing > Shirts & Tops > T-Shirts" is machine-readable. "Tops" is not.
When you invent your own category names, you're essentially speaking a dialect no AI system was trained to understand. Use the official taxonomy exactly. Find the most specific match available - depth matters more than you'd expect.
3. Description (First 160 Characters)
AI systems extract the first 160 characters of your product description to generate product summaries. That's about two sentences. What you put there determines what gets surfaced when an AI assistant summarizes your product to a potential buyer.
Write the first sentence as a complete product definition: who it's for, what it does, its key specs. "Lightweight trail running shoe for men with waterproof upper, carbon fiber plate, and 4mm drop. Built for technical terrain up to marathon distance." That's 160 characters that answer the questions AI systems are trained to answer. "Our best-selling shoe is finally here!" answers nothing.
4. Availability
This field must be real-time and must use Google's exact values: "in stock", "out of stock", "preorder", or "backorder". That's it. Those four. Custom values like "Available", "Yes", or "Ships in 3-5 days" won't parse correctly in most AI systems.
An AI assistant that recommends a product and sends a buyer to an out-of-stock page loses trust immediately. Platforms know this. Stale or non-standard availability data lowers your recommendation rate. Sync your feed at minimum once per day. Real-time sync via Shopify's product feed API is the right call if inventory moves fast.
5. Price (With Currency Code)
Include the currency code explicitly in your feed - USD, EUR, GBP. Not just the number. Some AI systems that receive price without a currency code default to displaying it without currency context. That creates bad user experience and lower recommendation rates because the platform can't confidently surface your pricing to buyers in other markets.
The correct format per Google Merchant Center spec is the number followed by the ISO 4217 currency code: "29.99 USD". A missing currency code is a small error with a compounding cost every time your product is evaluated for a cross-border recommendation.
6. GTIN/MPN
The Global Trade Item Number is how AI platforms match your product listing to product knowledge panels - the shared product data graph that aggregates reviews, pricing, and availability across merchants. Stores with GTINs get matched to that graph. Stores without them compete on title alone.
If you manufacture your own products and don't have a GTIN, use your Manufacturer Part Number (MPN) instead. If you resell branded products, GTINs are non-negotiable. Google requires GTINs for all products with assigned GTINs - and ChatGPT Shopping's product matching uses the same underlying product graph data. Missing GTINs mean your products stay invisible in product knowledge panels regardless of how good the rest of your feed looks.
7. Condition
Three values: "new", "used", "refurbished". That's the entire field. It has to be one of those three, stated explicitly. AI shopping systems filter recommendations by condition when a buyer's query includes condition signals - "new running shoes," "refurbished iPhone."
Missing this field doesn't mean AI platforms ignore it. Some treat a missing condition value as ambiguous. Others default it to "used." Either outcome is wrong for a new product. This is a five-second fix that removes a quiet filter blocking your products from new-condition queries.
8. Brand
Use your brand name exactly as it appears on your website, your social profiles, and your Google Business Profile. Not a variation. Not an abbreviation. Exactly the same string.
AI systems are building authority graphs that aggregate brand signals across platforms. When "WRKNG Digital" appears in your feed, "Wrkng Digital" appears on your Instagram, and "wrkngdigital" appears in your meta tags, those three strings don't automatically resolve to the same entity. Inconsistent brand names fragment your authority signal. Consistent brand names compound it. Audit every surface where your brand name appears and make them match.
9. Images (Minimum 800x800px, White Background)
Visual AI matching is live in both ChatGPT Shopping and Google AI Mode. The minimum viable image is 800x800 pixels on a white or neutral background with the product centered and unobstructed. Lifestyle-only photos - product in use, on a model, in a scene - perform worse in visual recommendation contexts because the AI's object detection has to work harder to isolate the product.
You don't have to eliminate lifestyle photos from your product page. You have to include at least one clean product shot that meets the spec in your feed image field. Google's Merchant Center guidelines specify image requirements in detail - 800x800px is the floor, 1000x1000px is better. Blurry, low-resolution, or heavily watermarked images are filtered out before they reach any recommendation surface.
Frequently Asked Questions
Q: Does fixing these fields guarantee my products get recommended by AI assistants?
No. These fields are table stakes - the minimum required to be considered. Getting them right makes you eligible. It doesn't make you automatic. But getting them wrong makes you ineligible, and that's the problem most Shopify stores have right now.
Q: How often should I update my product feed?
Availability and price should sync at least once daily. If your inventory moves fast, real-time sync via API is worth the setup time. Static fields like title, description, brand, and GTIN only need updating when the product information actually changes.
Q: Which of these nine fields matters most?
GTIN is the highest-use fix for most stores reselling branded products because it unlocks product knowledge panel matching. For stores with unique or private-label products, title structure is usually where the most recommendation signal is being left on the table.
Q: My Shopify store uses a third-party feed app. Does that handle all of this?
Most feed apps handle format correctly but don't fix data quality problems. They'll export whatever data is in your Shopify product fields - if your descriptions start with "Our best-selling product," that's what goes in the feed. The fields above are a data quality problem, not a format problem.
Q: How do I know which of these fields my current feed is getting wrong?
Google Merchant Center's diagnostics tab shows feed errors and warnings by field. That's the fastest free audit available. For a deeper look at how your feed performs specifically against AI shopping criteria, that's exactly what we do at WRKNG Digital.
If you want to know exactly where your product feed stands, get a free product feed audit and we'll show you which fields are costing you recommendations right now.
