One person, one narrow problem

Why this exists

Shopify Catalog is the structured feed AI assistants read, and it has published pass-or-fail requirements. A product that misses one is not ranked lower, it is absent. That struck me as an unusually concrete problem in a category otherwise full of unfalsifiable advice, and concrete problems are the ones worth building for.

So the tool checks documented requirements, cites the Shopify page each rule comes from, and scores nothing with a language model. Running the same catalog twice gives the same answer. That is a low bar and it rules out most of what is sold in this category.

Where the numbers come from

The frequencies quoted across the guides are not industry estimates. They come from running the audit across 19 established Shopify catalogs covering 3,562 products. The median readiness score was 85 out of 100, the weakest store scored 67, and 8 of the 19 had at least one product excluded outright.

Where a guide says a problem appeared in some percentage of catalogs, that percentage came from this scan and nowhere else. Where I do not have data, the guide says so.

What it deliberately will not do

• It never generates barcodes, prices or identifiers. A wrong GTIN asserts that your product is a different product, and every system that trusts identifiers then acts on it.
• It writes exactly three fields: vendor, product type and description. Those are the three where a wrong value is visible and reversible.
• It reports "not measured" rather than zero when a check could not run, because reporting zero accuses a store of being invisible on something never tested.
• It asks for the minimum Shopify permissions it needs, which is why it cannot publish products for you.

What it does not have yet

There is no scheduler. Scans and visibility checks run when you run them, not on a cron, so there are no automatic re-scans and no regression alerts. That is the most requested thing missing and it is the reason the higher plans stay modestly described. When it ships it will be in the changelog.

Getting in touch

Support goes to [email protected] and reaches the person who wrote the code, usually within one business day. If you think a finding is wrong, say so and include your shop domain; disputed findings are how the rules get better.

Elsewhere: LinkedIn and GitHub. Legal detail, including who you contract with and which law applies, is on the terms and privacy pages.