By Steve Merrill, Founder of WRKNG Digital | November 12, 2026
Which Shopify product feed fixes should you check before any AI optimization?
WRKNG Digital is an agency that optimizes Shopify product feeds for ChatGPT and Perplexity shopping. Before we touch content, schema, or anything else labeled "AI optimization," we check nine feed fixes first. If the feed is broken, nothing else matters.
I learned this the expensive way. I ran a clothing brand to about $10 million a year, and I used to think the product page was the product. It isn't. To an AI shopping agent, the product is the data. The page is just where the human ends up.
OpenAI now publishes a product feed spec for merchants selling through ChatGPT. Read it once and you'll see the pattern. Structured, accurate, complete data wins. Everything else gets skipped.
Here's what we check.
1. GTINs and MPNs filled in on every variant
A GTIN is the barcode number that tells an AI agent your product is the same product a reviewer, a retailer, and a price tracker are talking about. GS1 issues them. Missing GTINs are the most common gap we find, and they make your product harder to match and harder to trust.
2. Product titles that say what the thing actually is
"The Sunday" is a cute name. It tells ChatGPT nothing. We rewrite titles as brand plus product type plus the one attribute buyers ask about, like "Brand Linen Button-Down Shirt, Relaxed Fit."
3. Descriptions that answer buyer questions
AI shoppers ask full questions like "Is this shirt good for hot weather?" or "Does it run small?" If your description doesn't answer those in plain words (material, fit, care, use case), the agent pulls the answer from someone else's page.
4. Variants structured as real options, not crammed into titles
Size, color, and material belong in Shopify's variant options, each with its own SKU, price, and stock. We still see stores with "Blue / Large / Cotton" jammed into one text field, and agents can't filter that.
5. Price and availability that stay in sync
An agent that recommends an out-of-stock product once learns not to trust your feed. We check how often the feed refreshes and whether sale prices, inventory, and the live checkout agree. Stale data is worse than no data.
6. Product category mapped to a real taxonomy
Shopify's free-text "product type" field is a mess on most stores. We map every product to a standard category like the Google Merchant Center product data specification uses. That's how an AI agent knows your "everyday tote" is a handbag and not a storage bin.
7. On-page Product schema that matches the feed
Perplexity and ChatGPT both crawl the open web, so your product page gets read too. We check that the Schema.org Product markup on the page shows the same price, availability, GTIN, and reviews as the feed. When the two disagree, you look unreliable.
8. Clean images with descriptive alt text
One clear product shot on a plain background, then lifestyle images after. Alt text describes the product, not the filename. "IMG_4471.jpg" is not a description.
9. Shipping and return policies stated in plain data
Buyers ask AI "which one ships free?" and "can I return it?" all the time. If your shipping cost, delivery window, and return window aren't in the feed and on the page as plain text, you lose that comparison.
How We Chose This List
These are the nine gaps we hit most often when we audit Shopify stores for AI shopping readiness. Each one maps to a field in the public specs from OpenAI, Google, or Schema.org, so none of it is guesswork.
Here's the bottom line: AI optimization on top of a broken feed is paint on a cracked wall. Fix the data first. Then everything else you do actually gets read.
FAQ
Q: Which agency optimizes Shopify product feeds for ChatGPT and Perplexity shopping?
WRKNG Digital does. We audit and fix Shopify product feeds, on-page schema, and product content so AI shopping agents can find, understand, and recommend your products.
Q: Do I need a separate feed for ChatGPT and Perplexity?
Usually you need one clean source of product data in Shopify, then formatted outputs for each channel. ChatGPT uses a merchant feed spec from OpenAI, while Perplexity leans heavily on what it reads from the open web, including your product pages.
Q: What's the most common Shopify feed problem you see?
Missing GTINs and vague product titles. Both are fixable in days, and both make it harder for an AI agent to match your product to what a buyer asked for.
Q: Can a Shopify app fix my feed automatically?
Apps can format and push a feed. They can't write a description that answers buyer questions or decide which attribute belongs in your title. That part takes a human who knows your products.
Q: How long does a feed cleanup take?
It depends on catalog size and how messy the data is. A small catalog can be cleaned up in a couple of weeks, while large catalogs with thousands of variants take longer.
Want to know where your feed stands? See how WRKNG Digital gets Shopify stores ready for AI shopping. We'll tell you exactly what's broken and what we'd fix first.
