ChatGPT & GPT-4o Integration

ChatGPT is the highest-traffic AI shopping surface available today. When a user asks “What’s the best running shoe under £120?”, ChatGPT queries connected product sources before answering. Zevin registers your store as one of those sources.

How ChatGPT shopping works

OpenAI’s shopping layer works in two modes:

1. Browse with Bing — ChatGPT searches the web and surfaces products from indexed pages
2. Plugin / Action data sources — ChatGPT calls registered endpoints that return structured product data

Zevin targets both. Your UCP endpoint is indexed by Bing through Zevin’s sitemap infrastructure, and Zevin registers a ChatGPT Action that routes relevant product queries to your catalogue directly.

What shoppers see

When a shopper asks ChatGPT a product question relevant to your catalogue, they may see:

• A product card with your item’s image, name, price, and a direct link to your product page
• A comparison table when multiple items from your store match
• A “Buy at [your store name]” call to action that sends them to your checkout

Attribution is preserved — the session arrives at your store with a ?ref=zevin-chatgpt query parameter for analytics tracking.

Setup

1. Connect your store

Connect a Shopify, WooCommerce, or Magento store to Zevin. Your catalogue is automatically included in the ChatGPT feed.

2. Verify your UCP endpoint

In your Zevin dashboard, go to AI Platforms → ChatGPT and click Test connection. Zevin will simulate a ChatGPT product query against your live catalogue and show you the response.

3. Enable rich product cards

Rich cards (with images) require that your products have:

• At least one high-quality image (minimum 800×800px)
• A price in a supported currency
• A product description of at least 50 words

Zevin validates these requirements during sync and flags products that won’t render as rich cards.

GPT-4o and the Responses API

With OpenAI’s Responses API, developers can build custom shopping agents that query your Zevin-powered catalogue programmatically. Zevin exposes an OpenAI-compatible tool schema:

{
  "type": "function",
  "function": {
    "name": "search_products",
    "description": "Search the merchant's product catalogue",
    "parameters": {
      "type": "object",
      "properties": {
        "query": { "type": "string" },
        "max_price": { "type": "number" },
        "in_stock_only": { "type": "boolean" }
      },
      "required": ["query"]
    }
  }
}

Any GPT-4o agent can call this tool via your UCP endpoint with no additional configuration.

Coverage and latency

Metric	Value
Typical time to first ChatGPT appearance	24–48 hours after initial sync
Query latency (UCP → ChatGPT)	<200ms p99
Cache TTL	5 minutes for stock / price

Analytics

The Analytics dashboard shows ChatGPT-specific metrics including query volume, product impressions, click-through rate, and conversion.