Architecture · AI Commerce Intelligence

AI Commerce
Infrastructure

Before an AI agent can recommend or buy from your store, it moves through a stack of machine-readable layers, from a file at your domain root down to the checkout instructions. AI Commerce Infrastructure is that stack. It is the plumbing of agentic commerce.

What it is

The plumbing AI moves through

AI Commerce Infrastructure
AI Commerce Infrastructure is the full set of machine-readable layers a store exposes so AI agents can discover, understand, trust, and transact with it: llms.txt and llms-full.txt, agents.md, JSON-LD schema, a product feed, and crawlable policies.

Most store owners think about one layer at a time. They add schema, or they hear about llms.txt and turn it on. But an AI agent does not experience your store as separate features. It moves through them in sequence, and a gap in any layer stops it before it reaches the next. Infrastructure is the right way to think about it: a stack where every layer has to hold.

The stack

How AI moves through your store

From first contact at your domain root down to a completed purchase, here is the order an agent travels in and what lives at each layer.

1Discovery
llms.txtrobots.txt
At the domain root, the agent reads who you are and what you sell before crawling a single page. This is the first impression and it sets the frame for everything after.
2Catalogue
llms-full.txtproduct feed
The full inventory in structured form: every product, price, and category. The agent learns your complete range without crawling each product page one by one.
3Page
JSON-LD schemaserver-rendered HTML
On the individual product page, the agent confirms the facts: price, availability, rating, and brand as labelled structured data it can trust rather than guess.
4Transaction
agents.mdcrawlable policies
Finally, the checkout instructions and the return and shipping terms an autonomous agent needs to complete a purchase on a buyer's behalf.
A gap at any layer stops the agent before it reaches the next.
File reference

Every layer, explained

/llms.txt
A concise overview of your store at the domain root: positioning, key categories, trust summary. The first file AI reads. See the llms.txt guide.
/llms-full.txt
The complete version with your full catalogue and detailed product descriptions, used by agents that want to browse your entire range.
/agents.md
Checkout instructions for autonomous agents: how to add to cart, apply details, and complete a purchase on a buyer's behalf.
JSON-LD schema
Per-page structured data confirming product, price, rating, and identity. See AI Structured Signals.
product feed
A structured catalogue of everything you sell with prices and availability that must stay consistent with the live store.
policy pages
Returns, shipping, and contact as crawlable text pages, so the agent can read and verify the terms it needs before recommending you.
Shopify generates the first three for you. The Agentic Dashboard auto-creates llms.txt, llms-full.txt, and agents.md for every store. The work is not building them from scratch, it is making sure each layer is accurate, optimized, and consistent with the rest.
How to build it

Standing up the full stack

1
Optimize the discovery layer
Open your llms.txt and replace the generic auto-generated text with specific positioning, categories, price ranges, and a trust summary. This is the frame for everything below it.
2
Confirm the catalogue is complete
Check that llms-full.txt and your product feed reflect your real range, with prices and availability that match the store. Stale catalogue data quietly misleads every agent that reads it.
3
Layer schema on every page
Add and validate Product, Organization, and AggregateRating schema so the page layer hands agents confirmed facts instead of prose to interpret.
4
Make the transaction layer readable
Ensure agents.md is present and your return, shipping, and contact policies live on crawlable pages, so an agent can complete the journey rather than stalling at checkout.
FAQ

AI Commerce Infrastructure: common questions

What is AI Commerce Infrastructure?
It is the full set of machine-readable layers a store exposes so AI agents can discover, understand, trust, and transact with it. It includes llms.txt and llms-full.txt at the domain root, agents.md for checkout, JSON-LD schema on every page, a clean product feed, and crawlable policies. Each layer answers a different question an agent has on its way to a recommendation or purchase.
What is the difference between llms.txt, llms-full.txt, and agents.md?
llms.txt is a concise overview of your store that AI reads first. llms-full.txt is the complete version with your full catalogue and detailed descriptions. agents.md contains the checkout instructions an autonomous agent follows to complete a purchase. On Shopify the Agentic Dashboard generates all three automatically, though the auto-generated content usually needs optimizing.
Do I have to build all of this manually?
No. Shopify generates the llms files and agents.md automatically, and most themes provide base schema. The work is less about creating these layers from nothing and more about making sure each one is present, accurate, optimized, and consistent with the others, because a gap or contradiction in any layer weakens the whole stack.
Which layer should I fix first?
Start at the top. The discovery layer, your llms.txt, frames how every agent reads everything below it, and it is usually the most neglected. Then work down to schema and the transaction layer. A free AI Commerce Score shows which layers are weakest so you can prioritise.

Is your infrastructure complete?

The free AI Commerce Score checks each layer of your stack and shows where an agent would stall on the way to recommending you.

Free. No credit card. Results in 30 seconds.