parsance_

Shop platforms have started shipping AI readiness as a feature. Shopify generates an llms.txt with agent instructions automatically, catalogs have a standard machine-readable address, and more of this will come. Good. And in our July check of 20 German stores, 0 of 20 had shipping data in their schema and 6 of the 11 checkable product pages had no GTIN. Both things are true at once.

Update · July 18, 2026

Shopify has since proven the point at scale. The Spring ’26 Edition made agentic commerce the default: Catalog now syndicates eligible products into ChatGPT, Copilot, Google AI Mode and Gemini automatically, and checkout runs inside the chat. Access is no longer the question. Shopify’s own numbers say what is: catalog data converts at twice the rate of scraped data, and AI-driven orders grew roughly 13× year over year in Q1. The toggle is now everyone’s. The data underneath still decides which store the answer picks.

What the platform ships

The platform can generate everything that follows from structure. A file that tells agents where things are. A catalog endpoint. A schema skeleton on every product page. In our sample this works, verified, out of the box. The pipes exist, and the platform lays them for free.

What it cannot ship

The platform cannot know your delivery times. It cannot invent GTINs for your products, decide your return window, or collect your review data into markup you never connected. Every one of those is a piece of data only the merchant has, and exactly those pieces are where the sample fails. 0 of 20 with machine-readable shipping answers. 8 of 11 checkable product pages without rating markup. That is where what the platform can automate ends.

The free scan shows this split for your own store, signal by signal, platform part and data part.

My reading

Here is the opinion part, marked as one. I think the toggle creates a feeling of done. A merchant sees the platform announce AI support, assumes the box is ticked, and moves on. The data work that would make the support mean something never lands on a to-do list, because no dashboard shows it missing. That is a thesis, not a measurement, and n = 20 cannot prove it. What the sample does show is the gap between laid pipes and filled data, in every store we checked.

Filling that gap is merchant work. Making it stay filled is the job our maintenance product took over when it launched in July 2026, which is why the scan still comes first. Measure, then decide who does the filling.

How does your store read to AI agents?

Check your store for free