AI agents don't browse the way humans do
A human landing on your product page sees your photography first. The lifestyle shot. The packaging. The color. They scroll to the description when they want specifics. They check reviews when they want confidence. The whole page works together to sell.
An AI shopping agent lands on your product page and reads five fields. It cannot see your photos. It does not read marketing copy the way a person does. It parses structured data: your product title, description, metafields, taxonomy category, and image alt text. If those five fields are complete and precise, the agent can classify your product and recommend it. If they're thin or vague, the agent moves to the next result.
ChatGPT, Perplexity, and Shopify's own Sidekick all work this way. So does Google AI Mode. The underlying mechanism is the same across every AI shopping surface: structured fields in, recommendations out.
Shopify enabled Agentic Storefronts by default for eligible merchants in March 2026. Every qualifying Shopify store now has a live endpoint at /api/mcp that AI agents use to read your catalog. You did not opt into this. Your product data is already being read. The question is whether it's readable.
The five fields agents actually read
Fix these in order. Each one builds on the one before it.
1. Product title
Your title is the first field an agent reads to classify your product. It needs to tell the agent what the product is, who it's for, and what makes it distinct — without requiring any prior knowledge of your brand.
The format that works: [Brand] + [Product Type] + [Key Differentiator]. Include size, variant, or material if it differentiates. Drop taglines, creative names, and internal SKU codes.
"The Ritual Serum"
"Vitamin C Brightening Serum 30ml — Sensitive Skin Formula"
"Weekend Tote"
"Canvas Weekend Tote Bag 40L — Water Resistant, Natural"
Keep your creative names in your marketing and on your storefront display. Your catalog title serves machines. The two can differ.
2. Product description
Agents parse descriptions for attributes customers ask about in natural language queries. "Show me a moisturiser for oily skin under $30." "Find a dog harness that fits a 25lb dog." The agent scans your description for skin type, weight range, size, material, and use case. If your description skips those specifics, you don't match the query.
A description that works for AI covers: what the product is, who it's for, what it does, key ingredients or materials, size or dimensions, and any important compatibility or exclusion. Aim for one grounded paragraph before any lifestyle copy. State the attributes plainly. Marketing language can follow.
"Meet your new morning ritual. Our serum transforms your skin from the inside out. Glow like you mean it."
"15% Vitamin C serum formulated for sensitive and combination skin. Reduces dark spots and uneven tone in 4–6 weeks. 30ml. Fragrance-free. Suitable for daily use, morning and evening."
3. Metafields
Metafields store structured data that doesn't fit standard product fields: ingredients, certifications, compatibility, materials, care instructions, country of origin. Agents read these because they contain the specific attributes customers filter by in natural language queries.
The problem at most stores: metafield definitions exist, but values don't. A store sets up an "Ingredients" metafield definition six months ago. Fills it for the first 15 products. Never fills it for the next 60. Agents read those 60 products as having no ingredient data.
Go to Settings > Custom data > Products in your Shopify admin. Look at every definition. Then open your product list and filter by products where those fields are blank. Fill the gaps systematically, starting with your top 20 products by revenue.
Check before you define. Shopify's standard product taxonomy includes many built-in attributes. Before creating a custom metafield for something like "Material" or "Size," check whether the taxonomy category for your product already includes that attribute as a built-in field.
4. Product taxonomy
Taxonomy assigns your product to a category in Shopify's standard product hierarchy. The hierarchy has over 2,100 categories. Choosing the right one unlocks a set of category-specific attributes that feed Google Shopping and give AI agents structured attribute data to work with.
The difference between "Clothing" and "Apparel > Clothing > Tops > T-Shirts" is the set of attributes that unlock for that product: sleeve length, neckline, fabric, fit type, occasion. An agent matching a query for "slim-fit cotton crew-neck tee" reads those attributes. A product in the generic "Clothing" category has none of them.
Category: "Clothing"
Category: "Apparel > Clothing > Tops > T-Shirts" with attributes: Cotton, Slim Fit, Crew Neck, Casual
In Shopify admin, go to a product, scroll to "Product organization," click the Category field, and use the search to find the most specific category that fits. Shopify Magic will suggest categories. Accept or correct the suggestion, then fill in the unlocked attribute fields.
5. Image alt text
AI agents cannot see your product photos. They read alt text. A product image with no alt text is invisible to the agent. Descriptive alt text tells the agent what the image shows, which gives it context for the product's appearance, packaging, and usage.
Write alt text the same way you'd describe the image to someone on the phone. Include product name, color, material, size or scale cue, and shot angle for hero images. Be specific enough that someone who can't see the image would understand what's in it.
"" (blank)
"Vitamin C Brightening Serum in 30ml amber glass bottle with white dropper cap, shown against white background"
In Shopify admin, click into a product, click any image, and type alt text in the field that appears. Most stores have blank alt text on 80% or more of their product images. Start with hero images on your best sellers.
How to audit your store in under an hour
You don't need a developer or a specialist tool. Shopify's built-in export covers most of what you need to check.
Export your product catalog
Go to Products in your Shopify admin. Click Export. Select "All products" and export as CSV. Open the file in Google Sheets or Excel. You now have a row per variant and a column for every field Shopify tracks.
Sort by revenue, check the top 20
Sort by Shopify's "Total sales" or match your CSV against your sales data. Identify your top 20 products by revenue. Check those first. They're the ones AI agents are most likely to surface in queries, and the ones where gaps cost you the most.
Scan the five fields for each product
For each of your top 20, open the product in Shopify admin. Check: Does the title follow the [Brand + Type + Differentiator] format? Does the description open with specific, attribute-rich copy? Are all custom metafields filled? Is the taxonomy category specific? Does every image have alt text? Flag anything missing.
Fix gaps before moving on
Fix the fields for those 20 products before auditing the rest. Agents index your catalog continuously. Fixes apply on the next crawl, not on a schedule you control. Getting your top revenue products right first gives you the highest return for the time spent.
Time estimate: A catalog of 20 products takes 2–4 hours to audit and fix across all five fields. A catalog of 200 products takes 2–3 days. Budget accordingly, and start from the top of your revenue list every time.
Four mistakes that break AI discoverability
These show up in almost every catalog I review. Each one is fixable in a single afternoon.
Descriptive titles only on the homepage
Some stores use SEO-friendly product names on their homepage hero banners but keep short, creative titles in the actual product records. The agent reads the product record. Your homepage banner text is irrelevant to it. The title in Settings is the one that matters.
Metafields defined but never filled
An empty metafield definition tells the agent nothing. An agent reading a product with an "Ingredients" metafield that has no value treats it as missing ingredient data. Defining the structure is step one. Filling in values for every product is the actual work.
Inconsistent tags across the catalog
Tags like "organic," "Organic," and "organic-certified" on three different products read as three separate attributes. The agent sees no pattern. Pick one format per concept and apply it everywhere. Export your full tag list, spot duplicates, and standardize before you add more tags.
Boilerplate policy pages
AI agents read your shipping and return policy pages to answer customer questions: "Does this store ship to Canada?" "What's the return window?" A Shopify default template with placeholder brackets gives the agent an uncertain answer. Write specific, accurate policies. Agents reward clarity.
Clean data is the new storefront
For the first decade of Shopify, your product page served one audience: humans. Photos carried the weight. Copy set the mood. If a customer could see it and feel something, the page worked.
A second audience now reads your catalog. It reads five fields. It compares your fields against every competing product it can access. It recommends the one with the most complete, specific, consistent data. You don't see the comparison happening. You see the result: a sale, or no sale.
Stores that treat their catalog as infrastructure — data to be structured, not just copy to be written — will surface in AI-powered shopping. Those that don't will lose sales to competitors whose data is cleaner, and they won't know why.
Start with your top 20 products. Fix the five fields. Then work your way down the catalog.
See how your catalog scores across all five dimensions
The Structora ACR Score measures your Shopify store across Product Data Quality, AI Search Discoverability, Conversion Infrastructure, Ops Readiness, and Brand Consistency. Takes five minutes. No pitch.