Documentation

Every check below is derived from a document Shopify publishes, and each finding in the app links to the page it came from. Nothing here is scored by a language model, which is why the same catalog always produces the same result.

How scoring works

Findings fall into three levels, and only the first removes a product from the feed entirely.

The score weights blocking issues most heavily, because exclusion is binary. Your score is shown against the median of 85 from our scan of 19 established Shopify brands rather than against an abstract 100.

What is excluded from your score

Real catalogs contain entries that are not products: content blocks used to build landing pages, test rows, and zero-priced helper products created by bundle apps. Counting those as broken would be noise, so they are classified as housekeeping, left out of the score, and listed separately with the reason they were classified that way.

Zero-priced gifts and samples are a separate case. They are genuinely excluded from Shopify Catalog, so they are reported, but named as likely intentional rather than flagged as mistakes.

Blocking checks

A product failing any of these is absent from Shopify Catalog.

CheckFix
Product has no titleGive the product a descriptive title that names the item, not an internal code.
Product has no imageUpload at least one product image. One clear photo on a plain background is enough to qualify.
Product price is zeroSet a real price, or archive the product if it is a placeholder, sample or add-on that should not be sold on its own.
Product has no identifiable URLRestore the product handle so the item has a permanent storefront URL.
Product is not published to a storefrontPublish the product to your Online Store, Hydrogen or Headless channel.
Product status is UnlistedChange the status to Active if you want AI channels to be able to recommend this product.
Product is hidden from search enginesRemove the seo.hidden flag on products you want AI assistants to find.
Storefront is password protectedRemove the storefront password once you are ready to sell.

Matching checks

These fields are the ones Shopify names as considered by AI platforms when matching a product to a shopper request.

CheckFix
Product has no descriptionWrite a description covering what the product is, what it is made of, who it suits and when to use it.
Product type is not setSet a product type for every item, using consistent naming across the catalog.
Variant option name is genericRename variant options to what they represent, such as Colour, Size or Material.
No variant is in stockRestock, or archive the product if it is permanently discontinued so it stops diluting your catalog.
Barcode (GTIN, UPC, ISBN) is missingAdd the GTIN, UPC, EAN or ISBN to each variant.
Product belongs to no collectionAdd the product to at least one collection that matches how shoppers describe it.
Refund policy is missingPublish a refund policy in your store settings.

Minor checks

CheckFix
Description is too thin to match queriesExpand the description with materials, dimensions, care instructions and use cases.
Vendor is not setSet the vendor field, normally your own brand name or the manufacturer.
Product has no tagsAdd tags describing attributes shoppers actually ask for, such as material, occasion or fit.
Product has only one imageAdd at least one more image showing a different angle or the product in use.
Title is too short to disambiguateExtend the title with the defining attribute, for example the material, model or format.
Title is excessively longTrim the title to the product identity and move the extra detail into the description.
Privacy policy is missingPublish a privacy policy in your store settings.
Terms of service are missingPublish terms of service in your store settings.

What the app can write

Only vendor, product type and description, and only for changes you approve individually. Barcodes, prices and identifiers are never generated: those are facts about the world, and inventing them corrupts a catalog in a way that is expensive to detect later.

Every applied change records its previous value, so any batch reverts in one click from the Bulk fixes screen.

How AI visibility is measured

Buying questions are derived from your own catalog: your product types, price bands and brand. The set is then frozen and reused between runs, because a tracker whose questions change every time produces a trend line that means nothing. You can edit the set.

The headline metric is unbranded share of voice: the share of questions where an assistant named your store when the shopper never typed your brand. Questions that already contain your brand are reported separately, because being named after someone types your name proves nothing.

Category questions such as "what are the best coffee brands" have the same answer for every store in that category, so they are asked once and shared. Questions naming your brand are never cached or shared.

Permissions

PermissionUsed for
read_productsReading the catalog: titles, descriptions, variants, barcodes, collections, metafields
write_productsApplying bulk fixes you have approved, and reverting them
read_legal_policiesChecking whether your refund, privacy and terms pages exist, since assistants cite them

No customer permission is requested, so the app has no technical ability to read customers, orders or payment data. See the privacy policy.

More

Tutorial, FAQ, or email [email protected].