Case Study: How Velvet & Thread Doubled Their AI-Referred Revenue on BigCommerce

The Challenge

Velvet & Thread is a women's fashion boutique based in Austin, Texas, specializing in curated contemporary womenswear, accessories, and seasonal collections. They operate on BigCommerce with approximately 800 SKUs and generate around $600K in annual revenue. Their customer base is primarily style-conscious women aged 25 to 45 who value quality fabrics and unique designs.

In late 2025, the founder began testing how AI shopping assistants handled fashion queries. When she asked ChatGPT to recommend boutique women's fashion brands for spring, the results were dominated by large retailers and direct-to-consumer brands with massive marketing budgets. Velvet & Thread was nowhere. She tried Google AI Mode with queries like "best linen blazers under $200" and "sustainable women's fashion boutiques." Again, nothing.

The issue was not product quality or pricing. Velvet & Thread had strong organic search rankings and a loyal repeat customer base. The problem was that AI agents could not read, understand, or transact with their store. They ran the site through the AgentReadyHQ scanner in January 2026 and received an initial score of 24 out of 100 (F). Despite BigCommerce's built-in structured data capabilities, almost none of the AI-critical signals were in place.

Initial Audit Results

Here is the complete breakdown of Velvet & Thread's initial Agent Ready Score:

BigCommerce does provide built-in structured data for products, which gave Velvet & Thread a slight edge over platforms with no native schema support. However, the default output was missing critical fields for AI agents: no product variants with individual offers, no GTIN or MPN identifiers, no BuyAction markup, and no fashion-specific attributes like color, size, pattern, or material in the product feed. The Stripe integration was partially configured but had not been upgraded to support the Agentic Commerce Protocol.

Week 1 -- Foundation and Schema (Score: 24 → 55)

The first week focused on leveraging BigCommerce's existing infrastructure while filling the critical gaps that made the store invisible to AI agents.

1. Enhanced BigCommerce's Built-In Structured Data

BigCommerce generates basic Product schema automatically, but it omits several fields that AI agents rely on. Velvet & Thread used BigCommerce's custom fields feature to add structured attributes to every product. They created custom fields for brand, material, pattern, season, and care instructions. Then they modified their Stencil theme's product template to inject these fields into the JSON-LD output alongside BigCommerce's native schema. This approach preserved the platform's built-in structured data while extending it with fashion-specific properties.

2. Added GTIN and MPN to All Products

With 800 SKUs, this was manageable compared to larger catalogs. Velvet & Thread exported their product catalog via BigCommerce's CSV export, matched each product to its manufacturer's UPC or EAN code, and re-imported using the bulk product update tool. For their private-label items without standard GTINs, they used the MPN field with their internal identifiers and set the identifier_exists attribute to false in their product feed. This took roughly two days with one team member dedicated to the task.

3. Configured Stripe ACP for Agentic Checkout

Velvet & Thread was already processing payments through Stripe via BigCommerce. They upgraded to Stripe API version 2026-01 and enabled the Agentic Commerce Protocol features in their Stripe dashboard. This allows AI agents like ChatGPT to initiate secure checkout sessions where the customer confirms payment within the AI interface rather than navigating to the store. The integration required updating their BigCommerce checkout webhook handlers to support the new agent-initiated session flow.

4. Created Structured Policy Pages

Velvet & Thread had return and shipping information scattered across their FAQ and buried in footer text. They created dedicated, schema-marked pages:

• /shipping-policy with MerchantReturnPolicy schema and clear delivery timeframes
• /return-policy with a 30-day return window and free return shipping terms
• /size-guide with OfferShippingDetails schema for fashion-specific sizing data

Each page was linked from the site footer and from individual product pages, giving AI agents a clear path to parse the store's policies programmatically.

Week 2 -- Feed Optimization and Checkout SDK (Score: 55 → 78)

With the structured data foundation in place, week two targeted the product feed quality and BigCommerce's Checkout SDK to enable full agentic transactions.

1. Upgraded Google Merchant Center Feed with Fashion Attributes

Fashion is one of the most attribute-heavy product categories in Google Merchant Center. Velvet & Thread upgraded their feed to include every fashion-specific attribute that AI agents use for product matching and recommendations:

• color: Mapped from BigCommerce product options to GMC color values
• size: Full size ranges with size_type and size_system attributes
• pattern: Solid, striped, floral, plaid, and other pattern identifiers
• material: Primary fabric composition (linen, silk, cotton, etc.)
• age_group and gender: Required for apparel feeds
• product_highlight: Three key selling points per product

These attributes are what allow AI agents to match a query like "navy linen blazer size 8 under $200" to a specific product variant rather than just a product page. Without them, the store is invisible to attribute-filtered searches.

2. Enabled BigCommerce Checkout SDK for Agent Transactions

BigCommerce's Checkout SDK provides a headless checkout experience that AI agents can interact with programmatically. Velvet & Thread configured the SDK to expose a cart creation endpoint, a shipping rate calculator, and a payment initiation flow that works without requiring a browser session. This is a BigCommerce-specific advantage: the platform's API-first architecture makes headless checkout straightforward compared to platforms that require a rendered storefront for transactions.

3. Added BuyAction Schema to All Products

Beyond the Product schema, Velvet & Thread added BuyAction markup to every product page. This schema type explicitly tells AI agents that a product can be purchased directly, including the target URL for initiating a transaction and the accepted payment methods. They implemented this through a Stencil theme partial that dynamically generates BuyAction JSON-LD for each product, pulling the correct URL, price, and availability from BigCommerce's API.

Week 3 -- Technical Polish and AI Signals (Score: 78 → 91)

The final week focused on technical optimizations and emerging AI standards that push a store from "detectable" to "preferred" by AI agents.

1. Created LLMs.txt File

Velvet & Thread added an LLMs.txt file at their domain root that provides a machine-readable overview of their store for large language models. The file describes their product categories (dresses, tops, bottoms, outerwear, accessories), price ranges ($45 to $350), sizing information (XS through XL with petite and tall options), shipping regions, sustainability practices, and links to their structured data feeds. This emerging standard helps AI models understand what a store offers without crawling every page, and it is especially valuable for fashion where category boundaries and style descriptions matter.

2. Optimized Core Web Vitals

Velvet & Thread's site was loading in 5.2 seconds on mobile, primarily due to unoptimized product photography. Fashion stores are particularly susceptible to image bloat because high-quality lifestyle photography is central to the brand experience. They:

• Implemented automatic WebP conversion for all product images via BigCommerce's Akamai CDN
• Added responsive image srcsets with appropriate breakpoints for mobile, tablet, and desktop
• Deferred non-critical JavaScript including social proof widgets and chat plugins
• Enabled BigCommerce's built-in CDN edge caching for static assets

Mobile load time dropped to 2.1 seconds. AI agents factor site performance into their trust and reliability signals. A fast, stable store with accurate data gets recommended over a slow one with similar products.

3. Added FAQ Schema to Category Pages

Velvet & Thread added FAQ sections with proper schema markup to their top category pages. The questions were sourced from actual customer inquiries and common AI shopping queries in fashion:

• "What sizes do you carry?" with detailed sizing chart information
• "What is your return policy for sale items?" with clear terms
• "Do you offer free shipping?" with threshold and timeline details
• "What fabrics are used in your spring collection?" with material descriptions

FAQ schema gives AI agents ready-made answers for common questions, making the store more useful in conversational commerce contexts where customers ask natural language questions about product categories.

4. Implemented Real-Time Inventory via BigCommerce API

Fashion inventory moves fast, especially during seasonal transitions and sales. Velvet & Thread connected their BigCommerce store to Google Merchant Center using the BigCommerce GraphQL Storefront API for real-time product data. Instead of relying on daily feed refreshes, inventory changes now propagate within minutes. This prevents AI agents from recommending out-of-stock items or showing incorrect pricing during markdowns, both of which damage the store's trust score with AI systems.

Results After 30 Days

Velvet & Thread completed all optimizations by mid-January 2026. Here are the measurable results after 30 days of the changes being live: