Agentic storefront setup

When someone asks an assistant to find a linen shirt under a hundred pounds that ships to Germany, something has to answer. What it answers with depends on which stores it can read: which ones state their price, stock, materials and shipping in a form a machine can understand without guessing.

The substance is structured data and clear product information, which has been good practice for a decade. It used to buy you a slightly richer search result. Now it decides whether you're in the answer at all.

What assistants need to read

Complete, structured product data. Price, currency, availability, condition, a GTIN or MPN where one exists, and a product identifier that stays stable. Shopify provides a reasonable baseline. What's usually missing is the detail that decides a recommendation: materials, dimensions, fit, care, country of origin. If that only exists inside a paragraph of marketing copy, or inside an image, an assistant can't use it.

The fix is metafields, structured properly and written into the product's structured data as well as the page. On Nursem's store, for example, ingredients, routines and results live in structured fields the team edits in the Shopify admin, which is exactly the kind of data an assistant can read.

Accurate stock for every variant. "Is the shirt in stock?" has no useful answer. "Is the medium in navy in stock?" does. An assistant that recommends a sold-out variant gives a bad experience and learns not to trust the source, so availability has to be right for each variant, in the structured data itself.

Shipping and returns as data. "Where does it ship and what does it cost" is one of the most common questions, and on most stores the answer lives in a policy page written for people. Putting delivery destinations, costs and return windows into structured form is unglamorous and disproportionately useful.

Content that answers questions directly. Assistants pull out answers. Product and category pages that state the facts in plain sentences get used, and a page that's all brand atmosphere gives them nothing to quote. You can keep the brand voice; just make sure the facts are there too.

Letting assistants in

Assistants have to be able to fetch the page, and a few things commonly stop them.

Pages built in the browser. If product details only appear after JavaScript runs, some assistants see an empty page. Content rendered on the server is the safe position.

Over-eager bot protection. Rules tuned to block scrapers often block genuine assistant traffic too. Which agents you allow needs a deliberate decision.

Robots.txt and agent rules. What you allow is a commercial decision as much as a technical one. Some brands want to be in every assistant's index. Some don't want their catalogue used for training. Those are different questions, and they can have different answers.

What it's worth

Honestly, nobody has reliable numbers yet, and anyone quoting you a conversion uplift for agentic commerce is guessing. What I can say is that almost all of this work is worth doing anyway. Structured product data improves your search results and shopping feeds. Server-rendered pages load faster. Accurate stock for every variant cuts support emails.

That's the reasonable case for doing it now: it's cheap, it compounds, and most of the benefit doesn't depend on any prediction about AI being right. For more background, I've written about how AI shopping agents could reshape ecommerce.

Want to see what an assistant sees on your store today? The Shopify Plus audit checks it, or book a call.