Signal 1: llms.txt — a map of the store for AI
A machine-readable markdown file at the root of the domain: robots.txt, but for AI. Bots from OpenAI, Anthropic and Google already read it. Inside is a structured index: categories, key products, returns policies, FAQs. Without it, an agent burns tokens parsing JavaScript and frequently reads the store wrongly. With it, it gets a clean answer in milliseconds.
Since May 2026 Shopify generates llms.txt and agents.md automatically — every store has the file with no action from its owner. But auto-generation only gathers what is already in the admin: vague collection names and empty descriptions go into the index exactly as they are. Open yourstore.com/llms.txt and read it through an agent’s eyes.
Signal 2: extended JSON-LD schema
Basic Product schema with price and availability is the minimum viable. An agent needs the attributes that let it compare your product against a competitor’s: material_composition, weight, dimensions, warranty_details, energy_efficiency_rating, country_of_origin. Without them the agent sees your product as “one of ten faceless options” and picks on price. With them, you match specific queries like “1.2-litre cordless kettle, stainless steel, no plastic inside, at least a 2-year warranty”.
Signal 3: MCP access for agents
The Model Context Protocol is the handshake between an AI agent and a store. Through MCP an agent checks availability and price in real time and places an order — through a structured API rather than by parsing HTML. Shopify is already working on built-in support for Agent Access via MCP in Settings → AI & Automation. The announcements around agent payments through Visa Intelligent Commerce Connect make the point: an MCP-compatible checkout is becoming not an optional extra but a condition of entry into the agentic commerce funnel.
What to do this week
You do not have to do it all at once. You do have to start with the cheapest item.
Step 1. Open yourstore.com/llms.txt — five minutes. Check that the index shows your key collections and products and that the descriptions are not empty. If the file is not there yet, the feature is rolling out in stages — come back in a week.
Step 2. Open the Google Rich Results Test and check your five best-selling products. Find which schema attributes are missing. For a developer, that is 4–8 hours of work.
Step 3. Subscribe to Shopify’s announcements on agent access. When the feature ships, you will switch it on in a minute.
All of that costs one working day. The cost of doing nothing is that in a year an AI agent will buy from whoever did it already.