The Product Data Checklist for AI-Ready Fashion Brands

Fashion product data used to be treated as back-office plumbing. In agentic commerce it becomes part of the customer experience. A missing material field can remove a product from a constrained search. A stale price can break trust. An incorrect variant relationship can send a shopper to the right garment in the wrong size or color.

The goal is not to produce more text. It is to maintain one accurate product record that can be expressed consistently in a feed, structured data, platform catalog, product page, and checkout.

The 12-point checklist

1. Give every sellable variant a stable identity

Use a durable SKU or internal ID for each sellable variant. Where legitimate identifiers such as GTIN or MPN exist, preserve them exactly. Do not invent identifiers. A parent style ID should group variants, while each size-color combination keeps its own inventory and offer data.

2. Write a literal title before a poetic one

A useful product title identifies the item, defining style, material or construction when relevant, color, and variant information without promotional filler. “Bias-cut washable silk midi dress — cobalt” gives a system more usable facts than “The Midnight Muse.” Keep the evocative collection name elsewhere if it matters to the brand.

3. Choose the most specific accurate category

Category supports retrieval and comparison. A blazer should not live only under “new arrivals.” Map the internal category to the relevant commerce taxonomy and keep that mapping stable across seasons.

4. Describe the physical product, not an imagined customer

Record fiber content, lining, weight or hand where known, construction, closure, pockets, stretch, opacity, care, country of origin when verified, and included components. Avoid demographic or body assumptions that are not product facts.

5. Normalize color, pattern, and material

Brand color names can coexist with normalized values. Store “Midnight Current” as the marketing name and “navy blue” as the normalized color. Use the same controlled vocabulary across the catalog so filters and agents do not have to infer that “ink,” “deep ocean,” and “navy” may be related.

6. Make size and fit explicit

Record the size system, actual garment measurements where available, intended fit, model measurements when photography uses a model, and notable construction such as adjustable straps or elastic panels. Do not convert sizes across regions without a documented rule.

7. Synchronize price and availability

The visible product page, structured data, feed, and checkout should agree. Google’s Merchant Center documentation recommends matching the corresponding structured-data values and identifies price, priceCurrency, availability, and condition as important properties for automatic item updates. Automatic updates are a backstop, not a substitute for frequent feed updates.

8. Model shipping and returns as data

Agents and shoppers need more than “fast shipping.” Record eligible destinations, costs, handling time, transit estimates, pickup rules, minimum order conditions, and return windows in the platform fields designed for them. Make exclusions conspicuous.

9. Publish image sets with a job to do

Use a clear primary image, alternate views, material and construction details, and scale or fit context. Avoid embedding essential product facts only inside an image. Google’s 2026 Merchant Center update introduced an optional video_link attribute and announced that, beginning January 31, 2027, submitted product images will need a minimum resolution of 500 by 500 pixels across categories and marketing methods.

10. Label AI-generated product content where required

Google’s 2024 specification update added structured title and description attributes for generative-AI text and referenced IPTC metadata for AI-generated product images. Platform requirements continue to change. Maintain the original asset, generation record, disclosure metadata, and approval history so the same image is not represented as documentary product photography.

11. Map custom fields into the selling catalog

Shopify says its Catalog sends attributes such as title, description, options, images, price, and availability to agentic channels. When important data lives in metafields, metaobjects, tag prefixes, or custom title conventions, Shopify recommends Catalog Mapping so the correct values reach those channels.

12. Monitor disagreement, not just completeness

A catalog can be 100 percent filled and still be wrong. Run scheduled checks for:

• feed price versus landing-page price;
• feed availability versus selectable variants;
• parent-child variant consistency;
• broken image and product URLs;
• missing shipping or return terms;
• changes in platform warnings and disapprovals;
• AI-channel results that display an obsolete fact.

A minimum fashion product record

For a small brand, the following fields are a practical baseline:

id, parent_style_id, title, description, category, brand, product_url, primary_image, additional_images, price, currency, availability, condition, color, pattern, material, size, size_system, fit, measurements, care, shipping, returns, last_updated.

Not every channel accepts every field. The internal record should still be richer than any one export. Channel adapters can then map the source of truth into Google Merchant Center, Shopify Catalog, a direct merchant feed, schema.org markup, or a future commerce protocol.

Validation routine

1. Select one parent style with at least three variants.
2. Export its feed records.
3. Inspect the rendered Product and Offer structured data.
4. Compare both with the visible product page and checkout.
5. Change one test variant’s availability in the source system.
6. Record how long each destination takes to update.
7. Restore the original state and verify again.

This simple test reveals whether the catalog is truly synchronized or merely looks complete in the admin.

The principle to keep

AI readiness is a side effect of disciplined commerce operations. The same work that helps an agent understand a product also improves filters, internal search, customer service, ad feeds, accessibility, analytics, and marketplace listings.

Build the product record for truth first. Distribution follows.

Sources and verification

• Google Merchant Center: Product data specification, accessed August 31, 2026.
• Google Merchant Center: Supported structured data attributes and values, accessed August 31, 2026.
• Google Merchant Center: 2026 product data specification update, April 14, 2026.
• Google Merchant Center: 2024 product data specification update, April 9, 2024.
• Shopify: Catalog and product discovery for agentic storefronts, accessed August 31, 2026.
• OpenAI: Powering Product Discovery in ChatGPT, March 24, 2026.