By Steve Merrill, Founder of WRKNG Digital — November 3, 2026
How do I know if ChatGPT can actually see my Shopify products?
Ask it. Open ChatGPT, Perplexity, and Google AI Mode, type the exact question a buyer would ask, and watch whether your store shows up or a competitor does.
That is the whole test in one sentence. Most store owners never run it.
We wrote our guide to optimizing your Shopify product feed to cover the fixes. This post is the other half. The diagnostic. How to prove whether any of those fixes actually landed.
Because here is the trap. You can add schema, fill in every field, and still be invisible. The only way to know is to test the assistants the way a customer would.
I've run this on client stores where the feed looked perfect and the product still didn't surface. The data doesn't lie. So let's test yours.
What does the first test look like?
Prompt the assistants with a real buying question. Not your product name. The problem your product solves.
Say you sell merino wool base layers. Don't type your brand. Type "best lightweight base layer for cold weather hiking under $80." Run that same prompt in ChatGPT, in Perplexity, and in Google AI Mode.
Write down three things. Did your store appear? Which stores did appear? And did the assistant link a source or just describe products?
If a competitor shows up and you don't, that gap is your work order. The assistant found enough structured, matchable data about them to recommend them. It didn't find that about you.
Why should I test with and without the product name?
Because the two queries reveal two different problems. Named means brand recall. Described means attribute matching. You need both, and most stores only have one.
Run the query with your exact product title first. "Does [your product] come in a waterproof version?" If the assistant answers accurately, your product pages are getting read.
Now strip the name out. Ask for the same thing by description only. "Waterproof trail runner with a wide toe box under $130." If you vanish, the model can't connect a described need to your product.
That is a feed problem. The fields that carry material, fit, color, and use case are empty or vague. Buyers in AI chat describe needs. They don't type SKUs.
How do I check my structured data?
Use Google's Rich Results Test. Paste in a product URL and read what it says in about ten seconds.
You want valid Product markup showing name, brand, price, availability, and review ratings. Green means Google can parse it. Errors or warnings mean the assistant is reading a broken label and may skip you.
Check the required and recommended properties against Schema.org's Product type. The ones that move the needle for shopping answers are GTIN, brand, offers with price and availability, and aggregateRating.
One client passed the test but had no GTIN on any variant. AI shopping surfaces lean hard on GTIN to match a product across sources. We added it and their surface rate in Perplexity climbed inside two weeks.
How do I see which feed fields are actually exposed?
Open your product page source and look at the JSON-LD. Then open your Google Merchant Center feed and look at the same product. Compare what's filled against what's blank.
Shopify's own product fields map to attributes assistants care about. Title, description, type, vendor, and the variant options for size, color, and material. If those are thin on your end, they're thin everywhere downstream.
Make a short checklist and go field by field:
- GTIN or barcode on every variant
- Brand filled, not left as the store name by accident
- Color, size, and material as real variant options, not buried in the description
- Price and availability current
- A description that names the use case, not just the features
Blank fields are the silent killer. The assistant can only match what you give it. Give it nothing on fit and it can't answer a fit question with your product.
How often should I re-run these tests?
Weekly. AI answers aren't stable. The same prompt can return different stores on different days, and a model update can drop you overnight.
Run your core buying prompts once a week and log the result in a simple sheet. Date, prompt, did you appear, did they cite you, which competitors showed. Over a month you'll see a real trend instead of a lucky screenshot.
This is how you prove a fix worked. You add a GTIN, you wait, you re-run the prompt, you check the log. No guessing.
What if I pass every test and still don't show up?
Then the problem is authority, not formatting. The assistant can read you fine. It just trusts other sources more.
That usually means thin reviews, few third-party mentions, or no retailer listings carrying your GTIN. AI shopping answers pull from the open web, not only your site. Reviews on your product, mentions on comparison pages, and consistent product data across every place you sell all feed the model's confidence.
Fix the readable data first. It's free and it's fast. Then work on the trust signals. One is a technical problem. The other takes time.
One quick way to tell them apart. If the assistant describes your product correctly but recommends someone else, that's trust. If it describes your product wrong or not at all, that's still a data problem. Re-run the Rich Results Test before you blame authority. I've watched owners spend months chasing reviews when a single broken schema tag was the real block.
Frequently asked questions
Can ChatGPT read my Shopify product feed directly?
Not your raw feed. ChatGPT and Perplexity read your public product pages, your structured data, and third-party sources like reviews and retailer listings. If your Product schema is missing or thin, they guess from page text, and guesses lose.
What's the fastest way to test if AI assistants surface my products?
Ask the exact question a buyer would ask in ChatGPT, Perplexity, and Google AI Mode. If your store doesn't appear and a competitor does, you've found your problem. Thirty seconds, no tools.
Why does my product show up when I name it but not when I describe it?
The assistant can find you by brand recall but can't match you on attributes. Your feed is missing the fields (material, color, size, use case) that let a model connect a described need to your product.
What structured data do AI shopping assistants need?
Valid Product schema with name, brand, GTIN, price, availability, and aggregateRating. Google's Rich Results Test tells you in seconds whether yours is valid or broken.
How long until a feed fix shows up in AI answers?
In our client work, usually one to three weeks once the markup is valid and re-crawled. That's why you log the prompts weekly. You'll see the change when it happens.
Where to go from here
Run the five tests this week. Prompt the assistants, test named versus described, check your schema, inspect your fields, and start a log.
If you want help turning the gaps into a fix plan, that's what we do all day. See how we get Shopify stores surfaced in AI shopping at WRKNG Digital's agentic commerce page.
The stores winning in AI chat aren't the biggest. They're the ones that tested, found the blank fields, and filled them. Start testing.
