FoundRateWhy AI Shopping Assistants Skip Your Shopify Products: 9 Data Checks That Decide It
When ChatGPT answers "what should I buy?", it recommends products it can read, verify, and match to the question. Products it cannot read do not get argued against. They simply never come up.
That is the part most merchants miss. Losing in AI shopping rarely feels like losing. Your store does not get criticized; it gets silently skipped while a competitor with duller products but cleaner data takes the recommendation.
At FoundRate we score every product we scan against nine data checks. They are weighted by how much they affect whether an AI assistant can find, understand, and confidently recommend a product. Here are all nine, why each one matters, and the exact fix in Shopify admin.
1. A substantive description (the biggest factor)
The description is the AI's primary evidence about your product. Two lines of vibes ("Great mug! You'll love it") give it nothing to match against a real question like "handmade ceramic mug around 350 ml, dishwasher safe". We check for at least 200 characters of substance after stripping formatting.
Fix: rewrite the description to state what the product is, what it is made of, its size or dimensions, and who or what it is best for. Plain sentences beat keyword lists.
2. At least three images
Photo count is a trust signal to the systems ranking your page, and to the shopper who eventually clicks through. One lonely photo reads as an abandoned listing.
Fix: add at least three images, including one in-context or lifestyle shot. Give every image descriptive alt text; machines read the text, not the pixels.
3. A product type
The most common buying questions are category questions ("best gooseneck kettle"). Shopify's product type field is a direct category signal. Without it, your product is invisible to category-style matching.
Fix: in Shopify admin, under Product organization, set a specific type. "Gooseneck Kettle" beats "Kitchen" beats nothing.
4. At least three descriptive tags
Tags feed Shopify search, collections, and the feeds that machine systems read. They cheaply encode attributes your description might phrase differently: material, use case, style.
Fix: add three or more tags that describe the product's material, use case, and style, e.g. "ceramic, pour-over, handmade".
5. A vendor or brand
AI assistants cite products by brand. A product with no vendor set is hard to attribute, and an unattributable product is a risky recommendation.
Fix: set the Vendor field on every product, even if the brand is just your store's name.
6. A real, consistent price
Budget questions ("under $50") match on price. No price, a zero price, or one listing whose variants run from $8 to $800 cannot be matched to a budget with confidence, so it gets skipped.
Fix: put a price on every variant. If variants differ wildly in price, split them into separate products.
7. Material, size, and dimensions somewhere readable
Shoppers ask assistants practical questions: will it fit, what is it made of, how big is it. If those answers exist only in a photo or nowhere at all, the AI cannot answer them with your product, but it can with a competitor's.
Fix: state material and dimensions in the description, or add them as metafields. Text the machine can read is what counts.
8. Product structured data (JSON-LD) on your storefront
Product JSON-LD states your name, brand, price, and availability as machine-readable facts. It is the closest thing to speaking the AI's native language. Most Shopify themes include it, but some do not, and some apps break it.
Fix: open a product page, view source, and search for "@type": "Product". If it is missing, fix your theme or add a JSON-LD snippet to the product template.
9. In stock
Assistants avoid sending shoppers to dead ends. A product that looks unavailable is a bad answer, so it does not become the answer.
Fix: restock it, or hide it from the catalog until it is back rather than leaving a permanent "sold out" page.
How to use this list
Do not try to perfect your whole catalog at once. Pick your five most important products, run them against the nine checks, and fix what fails, starting with the description. Five products with complete data beat fifty products with thin data, because AI recommendations concentrate on the few products they can fully verify.
Then give it time
AI systems re-check stores every few days to a few weeks, not instantly. Apply the fixes, wait about a week, then re-test your buying questions. Same-day re-checks almost always show the old answer, and that is normal.
FoundRate runs these nine checks on your actual catalog, asks ChatGPT real buyer questions, and gives you a prioritized fix list in plain English. Run your free scan.