Shopify Product Feed API: Endpoints, Pagination, and What AI Agents Use

Three ways to access Shopify product data via API

Shopify provides three distinct APIs for reading product catalog data. Each serves a different audience and requires different credentials:

For AI agent discoverability, only the first one matters. /products.json is the endpoint external crawlers can reach without credentials. If that endpoint is broken or blocked, no amount of Admin API configuration will fix your AI visibility problem.

The /products.json API (what AI agents actually use)

Every Shopify store exposes its catalog at https://yourdomain.com/products.json. No API key, no OAuth, no Shopify account — just an HTTP GET request.

Basic request

curl "https://allbirds.com/products.json?limit=5"

The response is a JSON object with a single products array. Each item includes:

• id, handle, title — product identity
• vendor, product_type, tags — classification signals AI agents use for category matching
• variants — array of size/color combinations with price, SKU, and availability
• images — array of image URLs
• body_html — product description (HTML)

What it does NOT include: GTINs/barcodes, metafields, SEO meta titles/descriptions, or inventory quantities. For those fields you need the Admin REST API — but external AI agents cannot reach the Admin API, so those fields must be embedded in your Product JSON-LD markup on each product page instead.

Pagination

The default page size is 250 products. For stores with larger catalogs:

# Page 1
curl "https://yourdomain.com/products.json?limit=250&page=1"

# Page 2
curl "https://yourdomain.com/products.json?limit=250&page=2"

# Keep incrementing until products array is empty

Most AI shopping agents only fetch page 1. If your store has more than 250 products, only your first page gets bulk-indexed. Your sitemap (/sitemap.xml) is the secondary mechanism for agents to discover individual product pages beyond the first 250.

Filtering parameters

# Filter by vendor
curl "https://yourdomain.com/products.json?vendor=Nike"

# Filter by product type
curl "https://yourdomain.com/products.json?product_type=Sneakers"

# Filter by collection handle (via collection endpoint)
curl "https://yourdomain.com/collections/sale/products.json"

AI agents typically ignore filter parameters and fetch the unfiltered feed from /products.json or /products.json?page=N. Collection-specific endpoints (/collections/{handle}/products.json) are useful for your own integrations when you need a subset of the catalog.

Admin REST API (for your own integrations)

The Shopify Admin Products REST API lives at https://{shop}.myshopify.com/admin/api/2024-01/products.json. It requires an API key and access token — either from a private app or a Shopify Partners custom app with the read_products scope.

The Admin API returns more fields than /products.json, including:

• Metafields (including custom fields you define for AI agents)
• SEO title and description
• Product status (draft, active, archived)
• Barcode (GTIN) on each variant
• Inventory quantities across locations

Use the Admin API for bulk operations: pre-populating GTINs, rewriting descriptions with Claude API, or syncing data to your own database. Do not expose Admin API credentials in client-side code.

Storefront API (GraphQL, for headless storefronts)

The Shopify Storefront API is a GraphQL API intended for custom storefronts built with Hydrogen, Next.js, or other frameworks. It requires a storefront access token (a public, non-secret token).

If you have a headless store, your /products.json endpoint may return 404 because the request never reaches Shopify — it hits your Next.js or Vercel server instead, which has no route for it. The fix is to add a proxy route in your framework that fetches from the Storefront API and mirrors the /products.json response shape.

See our guide on the four signals you silently lose when you go headless for the complete proxy implementation for both Hydrogen and Next.js.

How to verify all three endpoints are working

# 1. Test public product feed (AI agent access)
curl -sI -A "GPTBot/1.0" https://yourdomain.com/products.json | head -3

# 2. Count products in feed (check for truncation)
curl -s "https://yourdomain.com/products.json?limit=250" | \
  node -e "const d=require('fs').readFileSync('/dev/stdin','utf8'); console.log(JSON.parse(d).products.length + ' products on page 1')"

# 3. Check if feed is referenced in sitemap (secondary discovery)
curl -s https://yourdomain.com/sitemap.xml | grep -c "products.json"

CatalogScan automates these checks and tests from multiple bot user-agents simultaneously, flagging cases where Cloudflare or a WAF treats bots and browsers differently.