By Steve Merrill · July 30, 2026
Does structured data actually help AI shopping recommendations?
Yes. Clean product structured data is how an AI assistant knows your price, your stock, your brand, and your rating. When those fields are right, your product shows up in the answer. When they're missing or wrong, you get skipped.
I've run this on client stores for two years now. The pattern is boring and consistent. The stores with clean Product schema get pulled into ChatGPT and Perplexity answers. The ones with broken or empty markup don't.
Here's the thing. Most stores treat schema like an SEO checkbox from 2018. It's not that anymore.
What structured data do AI shopping assistants actually read?
They read the Product type and the Offer inside it. That means name, brand, description, price, priceCurrency, availability, and an identifier like gtin or sku. Get those right and you've covered most of what matters.
The Schema.org Product docs list dozens of properties. You don't need dozens. An AI assistant answering "what's a good waterproof hiking boot under $150" needs a name, a price, a stock status, and something to tell one boot from another.
That's the shortlist. Nail it before you touch anything else.
Think about how the question gets answered. A shopper types a request into ChatGPT. The model has to pick a few products to name. It grabs the ones where it can state a price, confirm they're in stock, and tell them apart. Your schema is the source it grabs from.
If any of those fields is blank, you're a coin flip at best. Blank price. Blank brand. That's how you disappear from the answer.
The fields that carry the weight
- name, the exact product name, not a keyword-stuffed title.
- brand, AI uses this to group and compare. Missing brand is a silent killer.
- offers, price, priceCurrency, and availability. This is the field that decides if you get recommended when someone asks about budget.
- gtin / sku, the identifier that ties your product to a real thing in the world.
- description, a plain, honest description. AI reads it to match your product to the question.
What structured data can you skip?
Most of it. The long tail of Product properties adds noise and maintenance without moving recommendations. Depth, weight, and material only matter if a shopper actually asks about them for your category.
I've seen stores spend a week mapping 30 attributes into schema and get nothing for it. Meanwhile their price field was wrong. Fix the price first.
Skip the exotic stuff. hasMeasurement, additionalProperty, color variants nested five levels deep. Add those only when you've proven the basics work and you have a real reason.
One warning on reviews. Fake or inflated aggregateRating gets flagged. Google's own review snippet guidelines say the rating has to match what a person sees on the page. Mark up real reviews or none at all.
How do you set up structured data for AI shopping recommendations?
Start with the core Product fields, get the Offer right, add real reviews, then validate. Four steps, in that order. Don't jump ahead to the fancy markup before the basics are clean.
Step 1: Fix the core Product fields
Fill name, description, brand, sku, gtin, and price on every product page. Use JSON-LD. Google recommends it and it keeps your data separate from your page layout, which makes it easier to keep clean.
Step 2: Get the Offer and availability right
Add an Offer with price, priceCurrency, and availability. The availability value has to match your live stock. If your page says in stock and your feed says out of stock, the AI trusts neither.
Step 3: Add real review data
Mark up aggregateRating and individual reviews, but only with numbers a shopper can see. A 4.7 in your schema and a 3.9 on your page is a red flag that gets your whole markup discounted.
Step 4: Validate and keep the feed in sync
Test every template with Google's Rich Results Test. Then confirm your merchant feed and your on-page markup agree. AI shopping tools cross-check both. When they disagree, you lose.
Why do most stores still get this wrong?
Because the markup is right on 90% of pages and broken on the 10% that sell the most. A theme update, a sale price that never cleared, a variant with no gtin. Small gaps, big cost.
If you run Shopify, the platform generates a lot of this for you, but not all of it. Shopify's own SEO docs cover the basics, and plenty of stores stop there and assume they're covered. They're usually not.
The price and availability drift is the one that bites. Your checkout says one thing. Your schema says another. AI reads the mismatch and moves on to a competitor whose numbers line up.
Check your best sellers by hand. That's where the money is and that's where the breaks hide.
What does good structured data get you?
It gets your product into the shortlist an AI builds when someone asks for a recommendation. That's the whole game now. The assistant answers the question, and your product is one of the three it names.
Data doesn't lie about this. The stores that keep their Product schema clean and synced with their feed show up more often in AI answers. The ones that treat it as set-and-forget fade out.
Clean fields. Honest numbers. A feed that matches the page. That's what works.
I made the opposite mistake early on. Spent hours adding rich attributes to a client's catalog and never checked whether the prices were current. Half of them weren't. We fixed the prices in an afternoon and the AI mentions started coming back within two weeks.
So do the boring work first. It pays.
FAQ
Does structured data help AI shopping recommendations?
Yes. Clean Product schema gives AI assistants the price, availability, brand, and rating they need to recommend and compare products. Missing or wrong fields get your product skipped.
What structured data fields matter most for AI shopping?
Product name, brand, price, currency, availability, gtin or sku, and a real description. Ratings help too. These are the fields AI uses to answer buyer questions.
Do I need JSON-LD or is microdata fine?
Use JSON-LD. Google recommends it, it's easier to keep clean, and it keeps your data separate from your page layout. Microdata still parses, but JSON-LD is the safer default.
Will fake or inflated review markup get me recommended?
No, and it can hurt you. Ratings that don't match the page get flagged and ignored. Only mark up review data a shopper can actually see.
Want your products showing up in AI answers?
We audit your Product schema, fix the fields that matter, and get your feed and pages in sync. See how it works at WRKNG Digital's agentic commerce page.
