By Steve Merrill, Founder of WRKNG Digital — September 25, 2026
How do AI shopping agents pick which Shopify products to recommend?
They read your product feed and your structured data, then rank products by how complete and trustworthy the data is. A product with a full attribute set, a GTIN, live pricing, and visible reviews gets recommended. A product missing those fields gets skipped. The feed is the input. The recommendation is the output.
I've audited a lot of Shopify feeds this year. The same holes show up every time. Missing GTINs. Thin descriptions. Stale inventory. Zero review signals in the markup.
Agents like ChatGPT shopping, Perplexity, and Google's AI results don't guess. They match on fields. If the field is blank, you're invisible.
Here's how to fix your feed, field by field.
What product attributes do AI agents actually need?
Agents need the required attributes filled for every single variant: title, description, price, availability, condition, brand, GTIN, and MPN. When one is missing, Google Merchant Center flags the product and agents drop it from comparison. Completeness is the price of entry.
Start with the basics that most stores still get wrong.
- Title: Lead with brand, then product type, then the attribute that matters (color, size, material). "Stanley 40oz Quencher Tumbler, Rose Quartz" beats "Rose Quartz Tumbler."
- Condition: Set it. New, used, or refurbished. Agents filter on this and a blank field looks broken.
- Availability: in_stock, out_of_stock, or preorder. This has to match reality in real time.
- Price: Include currency and keep sale price and regular price both populated when you run a promo.
Google's product data specification lists every attribute and which ones are required by category. Read it once and map it against your feed. Google's Merchant Center product data spec is the source of truth here, not your gut.
Why do GTIN, brand, and category fields matter so much?
These three fields are how an agent identifies your product and compares it to the same item somewhere else. The GTIN is the barcode number that ties your listing to a global product record. Brand and category tell the agent what shelf your product sits on. Miss these and the agent can't place you.
GTIN. If your product has a manufacturer barcode, put it in. This is the single strongest identifier an agent uses to match your item across stores and pull in comparison data. No GTIN on a branded product means you lose the comparison entirely.
Selling handmade or custom goods with no barcode? Set identifier_exists to false so the feed doesn't error out, and make brand and MPN work harder.
Brand. Always populate it. Even for your own store label. Agents group and rank by brand, and a blank brand field reads as low trust.
Category. Map each product to a Google product category. Shopify lets you set this in the product taxonomy. Get it right and you show up in the correct comparison set. Get it wrong and you're competing against the wrong products.
How do you add structured data so agents can read your product pages?
Add Product schema to every product page using JSON-LD, including offers, price, availability, and aggregateRating. Agents scrape pages when they skip the feed, and structured data hands them the answer instead of making them guess from your HTML. This is your backup channel and it's free.
Shopify themes vary on how much schema they output. Most ship partial Product markup and leave out ratings. Check what your theme actually renders, then fill the gaps with an app or a theme edit.
At minimum, your Product schema needs:
name,description,brand, andskugtin13or the right GTIN length for your productofferswithprice,priceCurrency, andavailabilityaggregateRatingwithratingValueandreviewCount
Follow the schema.org Product type for the exact field names, and check Google's guidance on product structured data for what earns rich results. Valid markup that matches your visible page is the whole game. Mismatched data gets ignored or penalized.
What makes a product description work for AI extraction?
Descriptions work when they state real attributes in plain language: material, size, fit, use case, and specs. Agents extract answers from this text to respond to shopper questions, so vague marketing copy gives them nothing to pull. Write facts, not flourish.
Picture a shopper asking an agent "is this jacket waterproof and does it run small?" The agent answers from your description. If your copy says "elevate your adventure wardrobe," you lose. If it says "waterproof 3-layer shell, fits true to size, sized for layering," you win the recommendation.
Put the specs in the copy. Fabric weight. Dimensions. Battery life. Compatibility. The boring details are exactly what agents quote.
How should you handle images and availability?
Use clean, high-resolution product images on plain backgrounds, and keep availability and price synced in real time. Agents show images in their answers and won't recommend a product that's out of stock. Bad images and stale inventory quietly kill your listings.
On images: a clear main shot on white, multiple angles, and no promo text or watermarks stamped across the product. Agents and the feed both prefer clean visuals, and Merchant Center will disapprove images that are cluttered with overlays.
On availability: this is where I see stores bleed trust. An agent recommends your product, the shopper clicks, it's sold out. That store gets learned as unreliable. Shopify syncs inventory to Google automatically, but confirm the refresh is fast. Shopify's Google channel docs walk through connecting your store and checking sync status.
Do review signals change what AI agents recommend?
Yes, heavily. Agents need social proof to justify a recommendation, so products with visible star ratings and review counts get surfaced over identical products with none. Reviews are the tiebreaker when two products match on price and specs.
Expose your ratings two ways. Put aggregateRating in your Product schema so page scrapers see it. And connect a product review feed to Google Merchant Center so the feed carries ratings too.
We ran this on a client's store last quarter. Same products, same prices. Once ratings showed up in the markup, their products started appearing in AI comparison answers where they'd been absent before. Reviews were the missing input.
Where do most Shopify stores go wrong?
They treat the feed as a set-it-and-forget-it task. It's a living data asset. Prices change, inventory moves, new products ship, reviews come in. A feed that was clean in January is full of holes by September if nobody maintains it.
Run a monthly check. Pull your Merchant Center diagnostics, fix every disapproval, spot-check ten products against the field list above. That's an hour of work that decides whether agents can see you.
The stores winning in AI shopping right now aren't the biggest. They're the ones with the cleanest data.
Frequently asked questions
Do AI shopping agents use my Shopify product feed or my product pages?
Both. Agents pull structured feeds through Google Merchant Center and partner APIs, and they scrape product pages for structured data. Cover both so you're safe no matter which path an agent takes.
Is a GTIN required for AI shopping?
For branded products with a manufacturer barcode, yes. GTINs let agents match your item to the same product across stores and compare price and reviews. For handmade or custom items with no GTIN, set identifier_exists to false and lean on brand and MPN.
How often should I update my product feed?
Price and availability should sync in near real time. Shopify pushes updates to Google Merchant Center automatically, but check that your feed refreshes at least daily and that inventory changes hit within minutes, not hours.
Do reviews affect AI product recommendations?
Yes. Agents surface products with visible ratings and review counts because they need social proof to justify a recommendation. Products with no review data get passed over for competitors that show stars.
Get your feed ready for AI shopping
Your product data decides whether AI agents can recommend you. If your feed has holes, no amount of ad spend fixes it. Clean the data first.
Want help getting your Shopify store recommended by AI shopping agents? See how we do it at WRKNG Digital's agentic commerce page.
