-
Barcodes (GTIN) on every variant
The single highest-leverage field. A GTIN is how an agent matches your product to the same product elsewhere. Without one, your item cannot enter a comparison at all.
Where: Shopify admin → product → variant → Barcode. A wrong barcode is worse than none — it matches you to somebody else's product — so validate the check digit rather than pasting whatever the supplier sent.
-
Brand on every product
Fails every brand-named query when missing. One field, whole catalog.
Where: the product's Vendor field.
-
A real SKU on every variant
Variants without SKUs cannot be distinguished or reordered reliably. Partial coverage is its own problem: a product where three of five variants have SKUs reads as inconsistent data.
-
Shopify product category assigned
Decides which comparison you are eligible for. Shopify's Standard Product Taxonomy has over ten thousand categories, and "Home & Garden" is not a category — it is a refusal to pick one.
Be aware: Shopify exposes assigned categories on no public surface, so no external tool — including ours — can verify this one from outside. Anyone claiming to score it from your storefront is guessing.
-
Category attributes filled — the one that actually matters
Each category defines typed attributes: roast level for coffee, wax type for candles, sleeve length for shirts. These are what let a product survive a filtered request like "waterproof, size 11, under $150". Agents work through 30 or more such fields; typical stores expose five to eight.
This is the biggest single block of value and the one almost nobody has done. The first stores we measured scored 0.5% and 3.9% coverage.
Critical detail: every attribute has a closed list of allowed values. Write anything outside it and Shopify drops the value on import without telling you. The pattern attribute alone has 51 legal values, which is not something anyone guesses correctly.
-
Create the metafield definitions before importing
The failure that wastes the most time. If you import attribute columns for which no matching definition exists, Shopify reports the import as succeeded and silently discards those columns. The data is simply gone, and the success message tells you nothing is wrong.
Where: Settings → Custom data. Add the definitions first, then import.
-
Complete Product structured data on every product page
Not merely present — complete. Name, description, SKU, brand, GTIN, image, and an offer carrying price, currency, availability and condition. A partial block is a quiet failure: the page looks correct and the agent still cannot read the offer. Return policy and shipping timeframe belong here too.
Where: your theme's product template, or an app that injects it.
-
Named variant options
A product with multiple variants still labelled Title / Default Title gives an agent no way to honour "the large one in navy". Name the option Size, or Colour, or whatever it actually is.
-
Descriptions that answer a question
Not marketing atmosphere. If a shopper would ask "does this fit a 15-inch laptop", the answer should be in the text. Thin descriptions are where an agent gives up on intent it cannot resolve from typed fields.
-
llms.txt and agents.md
Two small text files at your domain root telling an agent where your catalog lives and how it may be used. Twenty minutes of work, and most stores have neither.
Check yours: yourstore.com/llms.txt and yourstore.com/agents.md. Also confirm your robots.txt is not blocking the agents you want — some CDNs add those blocks by default, without asking.
Scan your store against this checklist
The free scan measures items 1, 2, 3, 7, 8, 9 and 10 on your live storefront and names the products that fail. Items 4, 5 and 6 need Shopify's full taxonomy and are covered by the full report.