Your store has a new kind of visitor
In March and April 2026, Shopify shipped three releases that changed how people buy from your store. The short version: ChatGPT, Microsoft Copilot, Google AI Mode, and Perplexity can now browse your product catalog, build carts, and help customers complete purchases — no browser tab required, no app install required. A customer asks their AI assistant to find a product, and the assistant reads your store data to decide whether to recommend you.
You did not opt into this. Shopify turned it on by default.
Most of the coverage about these releases targets developers and agencies. This post is for store owners. You don't need to install anything or write code. You need to understand what changed, what these AI agents read when they visit your store, and what you can do about it today.
Three releases in three weeks
/api/mcp that AI shopping agents use to read your catalog, manage carts, and answer customer questions about your products and policies. This endpoint exists on your store right now.
AI-attributed orders on Shopify increased 11x between January and November 2025, according to TechCrunch. Some stores with strong AI visibility report 3–8% of their sessions coming from AI agent referrals. That number will grow. Shopify built the infrastructure for it.
Two parts, two audiences
The AI Toolkit has two components. Understanding the split helps you focus on what you control.
The Dev MCP (for your developer)
This connects AI coding tools to Shopify's documentation and API. Your developer or agency can use Claude Code, Cursor, or VS Code to build features, fix bugs, and manage your store with AI assistance. The AI coding tool reads Shopify's live documentation, so it produces accurate code on the first pass. Tasks that took a development sprint can now take a session. You don't interact with this layer. Your developer does. Ask them if they're using it.
The Storefront MCP (for AI shopping agents)
This is the part that changes your business. Every Shopify store now has a machine-readable endpoint that AI agents use to browse your products. A customer tells ChatGPT, "Find me an organic face serum under $40 with free shipping." ChatGPT reads the product data from every relevant Shopify store, compares options, and recommends one.
The agent reads your product titles, descriptions, tags, metafields, variant data, pricing, availability, and policy pages. If your data is incomplete, inconsistent, or vague, the agent skips you and recommends a competitor whose data is cleaner. You don't see this in your analytics. The customer never visited your site.
Sidekick, AI Toolkit, Agentic Storefronts: three layers
Shopify now runs three AI systems. They get confused with each other in press coverage. Here's the clear split:
Sidekick
The AI assistant inside your Shopify admin. It writes product descriptions, suggests campaigns, and answers questions about your store. You use this in your dashboard. It runs on Claude Sonnet 4.5. This is the merchant-facing tool.
AI Toolkit (Dev MCP)
Developer infrastructure. Your developer or agency uses this to build and modify your store with AI coding tools. You don't touch this. It makes your developer faster and more accurate — if your agency hasn't adopted it, ask them about it.
Agentic Storefronts (Storefront MCP)
The customer-facing surface. AI shopping agents from ChatGPT, Perplexity, Copilot, and Gemini use this to read your catalog and help customers buy from you. This runs on your store by default. You control it by controlling your product data.
Layer 3 is the one you need to act on. Layers 1 and 2 are tools. Layer 3 is a new sales channel.
AI agents are picky readers
A human browsing your store tolerates messy data — they look at photos, scan reviews, read between the lines. An AI agent reads structured fields. If the field is empty, the agent treats it as missing information. If the field is inconsistent with other products in your catalog, the agent loses confidence in your data.
Specific fields that agents read and use to make recommendations:
Descriptive titles rank. Creative titles don't.
"Vitamin C Brightening Serum 30ml" gives the agent three matchable attributes: ingredient, function, size. "The Glow Serum" gives the agent nothing. The agent cannot recommend a product it cannot classify.
Agents parse descriptions for attributes customers ask about.
A customer asks, "Show me a face serum for sensitive skin." The agent scans your description for skin type mentions. If your description says "formulated for sensitive skin," you match. If your description is three lines of marketing copy with no ingredient or skin type information, you don't.
Custom data fields are the strongest signal.
Metafields store structured data that doesn't fit in standard product fields: ingredients, compatibility, certifications, materials, care instructions. Agents read these because they contain the specific attributes customers filter by. Most stores have metafield definitions set up. Fewer than half fill them in for every product.
Category assignments determine which queries your products match.
Shopify's standard product taxonomy now has over 2,100 categories. Assigning "Shirt" is fine. Assigning "Apparel > Clothing > Tops > Shirts" unlocks attributes for size, neckline, sleeve length, fabric, and color — feeding Google Shopping data and giving AI agents the structured information they need to recommend you.
Consistent tags feed collection filters and agent queries.
If you tag one product "organic" and another "Organic" and a third "organic-certified," an agent reading your tags sees three unrelated attributes. Tag standardisation across your catalog determines whether your products surface in filtered searches — both on your store and through AI agents.
Agents can't see your photos. They read alt text.
A product image without alt text is invisible to AI agents. Descriptive alt text ("Vitamin C serum, 30ml amber glass bottle with dropper on white background") gives the agent visual context. Most stores leave alt text blank on 80% or more of their product images.
Six things you can do this week
You don't need a developer for any of these. You need time and attention to your product data.
Check your product taxonomy
Go to Products in your Shopify admin. Click into any product. Scroll to "Product organization" and check the Category field. If it says "Uncategorized" or shows a generic top-level category, assign a specific one. Shopify Magic will suggest categories. Do this for your top 20 products by revenue first.
Fill your metafields
Go to Settings > Custom data > Products. Look at which metafield definitions you've created. Then check how many products have values filled in. If you set up "Ingredients" six months ago but only filled it for 12 out of 80 products, those 68 products have a gap agents notice. Fill the gaps.
Rewrite vague product titles
Export your product list. Read every title. If a title requires prior knowledge of your brand to understand what the product is, rewrite it. Format: [Brand] + [Product Type] + [Key Differentiator]. Keep creative names in your marketing. Make your catalog titles machine-readable.
Add alt text to product images
In Shopify admin, click into a product. Click any image. Add alt text that describes what a person would see: product name, size, colour, packaging, angle. Do your hero images first. Work backward from your best sellers.
Standardise your tags
Export your products to CSV. Pull every unique tag into a column. Sort alphabetically. You'll find duplicates, inconsistencies, and abandoned tagging conventions. Build a tag dictionary. Pick one format per concept. Re-tag your catalog to match.
Review your policy pages
AI agents read your shipping policy, return policy, and FAQ pages. When a customer asks, "Does this store offer free returns?" the agent checks your policy pages. If your policies are boilerplate templates with placeholder text, the agent gives an uncertain answer. Write clear, specific policies.
Time estimate: Steps 1–6 take 2–4 days for a store with 50–200 products. Start with your top 20 revenue products — those are the ones AI agents are most likely to surface first.
This changes how you think about your catalog
For ten years, your Shopify catalog served one audience: humans browsing your website. Product pages were designed for eyes. Photos carried the weight. Copy set the mood. Tags and metafields were admin details most store owners filled in once and forgot about.
That model still matters. Humans still browse and buy. But a second audience now reads your catalog: machines. AI shopping agents that parse structured data, compare products across stores, and make purchase recommendations based on field completeness, data consistency, and attribute specificity.
Stores with clean, structured, complete product data will show up in AI-powered shopping. Stores with messy catalogs will lose sales to competitors whose data is easier for agents to read. You won't see these lost sales in your analytics. The customer never clicked through. They bought from the store the agent recommended.
Score your AI commerce readiness
The Structora ACR Audit scores your Shopify catalog across 5 dimensions: Product Data Quality, AI Search Discoverability, Conversion Infrastructure, Ops Readiness, and Brand Consistency. We walk through your store together and show you the specific gaps.