Preparing for the AI Shopping Era — What to Do for Shopify Agentic Commerce

Shopify has launched "Agentic Storefronts" — a system that enables AI agents like ChatGPT, Google Gemini, Microsoft Copilot, and Perplexity to directly discover, recommend, and purchase products from Shopify stores.

In this article, we dig into how AI agents actually read and process product data, drawing on technical sources, and outline what store owners should do to prepare.

AI Agents Select Products Using Structured Fields

Behind ChatGPT's shopping feature, there appears to be a protocol called the "Agentic Commerce Protocol (ACP)" defined by OpenAI. According to the technical specification, product data is synced every 15 minutes as structured feeds in CSV, TSV, or JSON format.

For Shopify stores, this product feed is likely auto-generated by "Shopify Catalog." The feed includes fields such as the following:

A Concrete Example

Suppose a user asks ChatGPT: "waterproof hiking backpack under $150." The AI agent would process this query as follows:

• category = backpacks → filter by category
• price < 150 → filter by price
• material metafield = waterproof material → match by attribute

If the "waterproof" attribute exists only in the description text rather than in a metafield, the product likely won't survive this filtering stage.

OpenAI's Agentic Commerce Protocol (ACP) technical documentation makes this point explicitly:

Two Stages of Product Data Processing — The Roles of Structured Fields and Descriptions

That said, descriptions are not entirely ignored. The ACP technical specification describes the role of product descriptions as follows:

Based on these insights, product data processing likely occurs in two stages:

1. Stage 1: Filtering by structured fields — Narrowing candidates using category, price, availability, and metafield values (material, size, etc.)
2. Stage 2: NLP analysis of descriptions — Evaluating how well each remaining candidate matches the user's intent through natural language processing

In other words, metafields act as a "qualifying gate" to remain in the candidate set, while descriptions serve as "bonus points" for being selected from within that set. If a product doesn't pass the gate, even the most compelling description will never be read.

How Shopify Catalog Works — What's Automatic and What's Manual

Shopify Catalog appears to be Shopify's central infrastructure for distributing product data to AI agents. The Shopify Help Center describes it as follows:

With this in mind, let's categorize each data type as "automatic," "semi-automatic," or "requires setup."

Automatic: Standard Fields

Title, description, price, inventory, images, and variant options appear to be automatically included in Shopify Catalog without any configuration. As long as your store has Agentic Storefronts enabled, this data is distributed to each AI platform.

Semi-Automatic: Category Metafields

Shopify has a "Standard Product Taxonomy" with over 10,000 nodes. When you assign a category to a product, category metafields associated with that category are automatically suggested.

For example, selecting "Apparel & Accessories > Clothing > Shirts" will suggest the following:

• Size, neckline, sleeve length
• Material (fabric)
• Target gender, age group
• Color, clothing features

These are standard metafield definitions under the shopify. namespace, and they appear to be automatically synced to Shopify Catalog as structured data. No Catalog Mapping configuration is required.

The Big Picture: What's Automatic vs. Configurable

Here's a summary of how product data flows to Shopify Catalog:

Below is an example of the category metafield definition screen. These are standard definitions under the shopify. namespace, with the associated categories also displayed. When you populate values in these fields, they appear to be automatically included in Shopify Catalog without any Catalog Mapping configuration.

Price, availability, images, variant options, and category metafields are all automatically ingested by Shopify Catalog, and their source cannot be changed (nor does it need to be). The only fields you can remap via Catalog Mapping are the three where stores might want to differentiate between "human-facing" and "AI-facing" content: product title, description, and category.

Requires Setup: Custom Metafields and Catalog Mapping

Custom metafields created by merchants (e.g., custom.specs, custom.materials) are not passed to AI agents by default. The Shopify Help Center provides the following guidance:

Catalog Mapping is configured in Shopify Admin under Settings → Agentic Storefronts → Catalog Mapping. The three fields available for mapping are:

While the UI presents all three fields side by side, category metafields are tied to the product category and are automatically included in the Catalog on a separate layer. Changing the category source risks breaking this linkage, so it's safest to only modify the "Product title" and "Product description" sources.

Additionally, stores on the Agentic plan can configure custom variant grouping (classifying variants by title, metafield, or tag prefix).

Does Metafield Completeness Directly Impact AI Visibility?

Shopify Growth Services' official guide states the following:

That said, the qualitative claim — that richer attributes improve AI discoverability — is logically supported by the underlying technical architecture. Here's why:

1. The filtering mechanism — AI agents filter by structured fields. Products with empty attributes fail to match filter conditions and drop out of the candidate pool.
2. Handling comparison queries — Queries like "which is lighter, A or B?" or "is it machine-washable?" require structured attribute values. Extracting this from description text is unreliable.
3. Product clustering in Shopify Catalog — The Catalog uses attribute data to classify and deduplicate products. Products with sparse attributes are less likely to be classified correctly.

What to Do Now — A Prioritized Action List

Agentic Storefronts' checkout integration is not yet available outside the US (as of March 2026, it's in US Early Access). However, AI agents discovering products through Shopify Catalog and web crawling is already happening regardless of region. Whether or not checkout is available, the following actions deliver immediate value.

Priority 1: Assign Accurate Product Categories

Assign the most specific category from Shopify's Standard Product Taxonomy to every product. Go beyond "Footwear" — drill down to the level of "Men's Insulated Winter Boots."

Shopify Magic auto-suggests categories based on product title, description, and images, so use those suggestions as a starting point and verify or adjust as needed.

Priority 2: Populate Category Metafield Values

Once a category is assigned, fill in as many of the suggested metafield values as possible for every product. This is the most direct way to clear the "qualifying gate."

Which fields to prioritize varies by category:

Priority 3: Create Separate Product Descriptions for AI Distribution

There's an important nuance here. Product descriptions are originally written for customers visiting the storefront. Rewriting brand storytelling and marketing copy into a dry spec list would defeat their purpose.

With Catalog Mapping, you can set up a separate AI-targeted description without touching the storefront description at all.

The mechanism is straightforward:

1. Create a custom metafield (e.g., custom.ai_description)
2. Populate it with an AI-optimized description for each product (a 150-300 word attribute-rich text covering use case, material, specs, target audience, and competitive advantages)
3. In Catalog Mapping, switch the "Product description" source to this metafield

This creates the following data split:

The same approach works for product titles. You can keep creative, brand-specific titles on the storefront while serving AI channels with a search-friendly format like "Brand Name + Product Type + Key Differentiator."

Priority 4: Configure Catalog Mapping

To actually deliver the AI-optimized data you prepared in Priority 3, you need to configure Catalog Mapping. Open "Shopify Catalog Mapping" from your Shopify Admin.

Step 1: Map Product Fields

The mapping screen displays three items: "Product title," "Product description," and "Product category." However, you should only modify "Product title" and "Product description."

1. Go to Shopify Admin → Shopify Catalog Mapping
2. In the Product fields section, click the dropdown for the field you want to change:
   • Product title → Select the metafield containing your AI-optimized title
   • Product description → Select the metafield containing your AI-optimized description
3. In the Product summary section, click "Change" to select a different product and verify the preview
4. Click Save

Step 2: Custom Variant Grouping (If Applicable)

This setting is for stores with complex variant structures. By default, Shopify uses the "Combined Listings" configuration, but you can set up custom grouping as follows:

1. In Catalog Mapping, activate Custom variant grouping
2. Choose a grouping method:
   • Product title — Group by delimiter characters in the title (- or |)
   • Product metafield — Group by a specified metafield value
   • Product tag — Group by tag prefix (e.g., group:summer-collection)
3. Verify the preview → Click Save

When custom variant grouping is enabled, you can also configure the display names for option labels (color, size, etc.) shown on AI channels.

Priority 5: Image Alt Text and GTINs

Add descriptive alt text to all product images (this serves as supplementary data for multimodal AI), and assign valid GTINs (UPC/EAN) to each variant.

Summary

Shopify has already built the framework for AI agents to properly understand your products. Select a category and metafields are suggested. Fill in the values and Shopify Catalog automatically distributes them to AI agents. If you want to keep human-facing descriptions separate from AI-facing structured data, Catalog Mapping lets you switch the source.

The key isn't cramming everything into product descriptions. It's putting the right data into the structured fields that Shopify has already prepared. The format for AI readability is already in place — structuring your product data to match it is the best place to start.

---

Sources

• DEV Community — OpenAI's Agentic Commerce Protocol: a technical look at how ChatGPT becomes a shopping agent
• Shopify Help Center — Shopify agentic storefronts
• Shopify Help Center — Mapping your product data sources for Shopify Catalog
• Shopify Growth Services — How to Prepare Your Shopify Store for Agentic Commerce
• Charle Agency — Shopify Product Taxonomy: The Complete Guide for 2026
• Shopify Help Center — Category metafields
• Shopify News — Introducing Shopify Agentic Storefronts