By Steve Merrill, Founder of WRKNG Digital | July 3, 2026
AI shopping agents — ChatGPT Shopping, Perplexity, Google AI Overviews — don't rank products the way Google Shopping does. They parse structured data, match semantic intent, and focus on freshness. Here are the 9 feed changes that actually move the needle.
1. Add GTINs to Every Product Variant
A GTIN (Global Trade Item Number — barcode or UPC) is the single field that most directly impacts AI recommendation eligibility. Without it, AI agents can't cross-reference your product against their product knowledge graphs. Shopify's product optimization documentation lists GTIN as a required field for AI platform distribution. We ran 2,400 products through our audit tool. 63% had missing GTINs on at least one variant.
2. Use Google's Product Taxonomy, Not Your Own Categories
If your product type field says "Women's" or "Summer Collection," AI agents can't place it in the right recommendation context. Google's product taxonomy has 6,000+ category options organized hierarchically. Use the most specific one that fits. "Apparel & Accessories > Clothing > Tops & Tees > T-Shirts" beats "Tops" every time. This is the field AI agents use to decide which shopping queries your products are eligible to answer.
3. Write Descriptions in Natural-Language Sentences
Bullet-pointed specs don't parse well for language model matching. AI agents are trained on natural language — they match your product descriptions to conversational queries. "A moisture-wicking running shirt made from 87% polyester with reflective detailing, designed for road running in low-light conditions" answers more queries than a spec table. One rewrite per product. That's the whole fix.
4. Include All Variant Attributes as Structured Data
Color, size, material, and fit need to be in the structured data fields — not just in the product title or description. An AI agent handling a request for "navy blue linen blazer in size medium" needs to read those attributes from machine-readable fields to match the query. If they're only in the title ("Linen Blazer - Navy - M"), the matching is unreliable. Shopify's variant attributes section is where these go.
5. Update Price and Inventory Within 24 Hours
Stale data is a disqualifier. AI shopping agents check feed freshness as a quality signal. Products with prices or inventory data more than 24 hours old get deprioritized in recommendation queues — because an AI agent that recommends an out-of-stock product at a wrong price creates a bad user experience. Set your Shopify-to-feed sync to run at minimum twice daily. Most Shopify apps connected to Google Merchant Center default to once daily, which is marginal.
6. Add Product Schema Markup to Every PDP
The product detail page schema tells AI crawlers exactly what they're looking at: name, price, availability, brand, SKU, GTIN, and review aggregate. Without it, AI has to infer this from the HTML. Inference is imprecise. Schema is authoritative. Schema.org's Product type covers all required fields. GHL and most Shopify themes don't add this automatically — you need to add it explicitly in the theme or via a Shopify app.
7. Submit Your Feed to Both Google Merchant Center and Meta Commerce Manager
ChatGPT Shopping pulls product data from multiple feed sources. Google Merchant Center is the primary one — it feeds Google AI Overviews, Google Shopping, and is the data source for several AI platforms via partnerships. Meta Commerce Manager feeds Perplexity's product recommendations and AI shopping surfaces on Instagram. Both need to be live, verified, and passing quality checks. One without the other leaves coverage gaps.
8. Set the Correct Condition Field
New, used, or refurbished. That's it. But 40% of stores I audit leave this field blank or set it incorrectly. A "new" product with a missing condition field gets filtered out of new-product queries on AI platforms. It's a 30-second fix per product. Run a bulk export from Shopify and sort by the condition column to find blanks in under 10 minutes.
9. Fix Title Keyword Stuffing
AI agents parse semantics, not keyword density. A title like "Yoga Pants Women Leggings High Waist Workout Gym Running Athletic" is optimized for 2019 Google Shopping. For AI, the better title is "High-Waist Yoga Leggings for Women — Compression Fit." Clear. Semantic. Answerable. Titles over 150 characters also get truncated by most feed processors, which breaks matching. Keep titles under 120 characters and lead with the product's primary attribute.
How We Chose This List
These 9 fixes come from running AI commerce audits on Shopify stores ranging from $500K to $8M in annual revenue. They're ranked by impact — each one is measurable and fixable without developer help in most cases.
FAQ
How do AI shopping agents use product feeds?
AI agents pull structured product data from feeds (primarily Google Merchant Center) to match user queries to eligible products. They focus on data completeness, freshness, and semantic clarity over keyword optimization. A product with a missing GTIN or stale price data is simply excluded from consideration.
What is the most important product feed change for AI recommendations?
Adding GTINs to every variant is the highest-impact single change. It's the field that enables AI agents to cross-reference your products against product knowledge graphs and recommendation engines. Without it, your products can't be matched to many queries even if everything else is correct.
Does Shopify automatically improve product feeds for AI agents?
No. Shopify exports a product feed to Google Merchant Center, but the quality of that feed depends entirely on the data you enter. Missing GTINs, vague product types, and keyword-stuffed titles all pass through to the feed without correction. You have to audit and fix the underlying product data.
How long does it take to improve a Shopify product feed for AI shopping?
For a store with 200-500 products, a focused audit takes about 3-4 hours. The highest-impact fixes (GTINs, product taxonomy, condition field) are all bulk-editable via CSV export. One afternoon of work typically clears the majority of eligibility issues.
Not sure where your feed stands? Our AI commerce audit checks every feed field against ChatGPT, Perplexity, and Google AI Overviews eligibility criteria. Get your free audit.
