Product feed management for marketplaces: A practical guide
ChatGPT Product Feed (2026): specification for ChatGPT Shopping, Agentic Commerce
The OpenAI Product Feed enables merchants to list and manage their catalogues within ChatGPT’s Shopping experiences. The feed acts as the single source of truth for all product data (titles, prices, stock, media, logistics) and powers search, discovery, and (where eligible) instant checkout. Product data are not “crawled”; instead, the merchant pushes a structured file to a secure OpenAI endpoint, with frequent updates ensuring accuracy. See how to connect your product feed to ChatGPT.
Table of content
Core principles
- Structured, merchant-controlled feed: ChatGPT indexes a feed provided directly by the merchant, ensuring accurate pricing and availability (no passive crawling).
- Accepted formats: CSV, TSV, XML, or JSON.
- Update frequency: refreshes are accepted as often as every 15 minutes to keep prices and stock nearly real-time.
- Instant Checkout: in-chat purchasing is powered by the Agentic Commerce Protocol (ACP), using integrations with PSPs such as Stripe for secure checkout.
Formats and delivery
- Formats: CSV, TSV, XML, or JSON (whichever suits existing catalogue exports best).
- Delivery: the merchant pushes the feed via encrypted HTTPS to an allow-listed endpoint provided by OpenAI (no pull model).
- Frequency: updates may be sent every 15 minutes to maintain fresh pricing and inventory data.
- Validation: an initial sample or full feed submission is used to confirm compliance before going live.
Required fields (minimum)
Each product must include the following attributes to qualify for indexing:
- id (unique, stable identifier)
- title
- description (plain text, no HTML)
- link (product page URL)
- image_link (primary product image URL)
- price (value plus ISO 4217 currency)
- availability (e.g. in_stock, out_of_stock, preorder)
- inventory_quantity
- brand
- condition (e.g. new, used, refurbished)
- enable_search (boolean - allows appearance in ChatGPT search results)
- enable_checkout (boolean - activates direct purchase; requires enable_search=true)
- Merchant policies for checkout: return_policy, seller_privacy_policy, seller_tos.
Recommended and enriching fields
- Product identifiers: GTIN, UPC, or MPN (improve catalogue matching and reduce errors).
- Logistics data: weight, dimensions, delivery regions and costs (shipping), estimated delivery times (delivery_estimate).
- Variants: item_group_id, colour, size, gender, size_system.
- Rich media: additional_image_link, video_link, model_3d_link (GLB/GLTF).
- Trust and signals: review counts, ratings, popularity data.
- Compliance: warnings, age restrictions, or safety details.
OpenAI-specific flags
- enable_search: determines whether the product appears in ChatGPT Shopping search results.
- enable_checkout: enables in-chat purchasing; automatically requires enable_search=true.
Best practices
- Accuracy and freshness: provide clear, complete, and up-to-date information - concise titles, factual descriptions, correct prices and stock levels.
- Semantic structure: avoid “marketing fluff”; use structured attributes and consistent taxonomy values.
- Automate updates: schedule exports every ≤15 minutes to prevent pricing or stock mismatches.
- Transparency: include seller_name and publish return, privacy, and terms policies to build shopper trust.
Agentic Commerce Protocol (ACP) and Instant Checkout
The Agentic Commerce Protocol is an open standard that manages secure communication between buyers, AI agents, and merchants to complete purchases, covering checkout, payments, fraud signals, and fulfilment notifications. The first live use case is Instant Checkout in ChatGPT, now rolling out to selected merchants and regions.
- Associated specs: Agentic Checkout Spec (checkout flow) and Delegated Payment Spec (secure payment sharing with PSPs such as Stripe).
- Security: TLS 1.2+, HTTPS on port 443, and documented privacy and compliance policies before production approval.
ChatGPT Shopping (search and recommendations)
When ChatGPT detects shopping intent, it can surface products with images, details, and direct purchase options. Results are not ads
and are not influenced by paid placement; they rely entirely on feed quality, product relevance, and user context.
Quick comparison: Google Shopping vs. OpenAI
- Similar structure (structured catalogue, required and optional attributes) but a distinct OpenAI schema with unique flags (enable_search, enable_checkout).
- Conversational focus: product descriptions are designed to perform well in natural-language search within ChatGPT.
- Merchant control: push model with high-frequency refreshes prevents outdated data and ensures accuracy.
Official links and useful resources
- OpenAI – Product Feed Specification
- OpenAI – Key Concepts (ACP and feed)
- OpenAI – Getting Started with Agentic Commerce
- OpenAI – Agentic Commerce in Production (security and compliance)
- Buy It in ChatGPT – Instant Checkout
- Help Centre – Shopping with ChatGPT Search
- OpenAI – Agentic Checkout Specification
- OpenAI – Delegated Payment Specification
- SEO.ai – Submit Your Product Feed to ChatGPT
- Backlinko – ChatGPT Shopping Optimisation
- Premier Octet – Agentic Commerce Guide (OpenAI x Stripe)
- eesel.ai – Agentic Commerce Protocol and Stripe
- OpenAI – Introducing AgentKit
Our guides
Guide
Case Study
10/09/2025
1 min read
Dr. Pierre Ricaud Case Study: +89% Profitability with Feeds
Guide
eBook
03/09/2025
1 min read
Boost Ecommerce Ads
Guide
Case Study
02/09/2025
1 min read
Darty Case Study: Live on TikTok Shop in 40 Days
Guide
eBook
13/06/2025
1 min read
Reseller Monitoring
Guide
Price Intelligence
31/07/2025
1 min read
Competitive Intelligence
List everywhere, manage it once
AI-powered feed management, global price monitoring, and the analytics to connect them, all in one platform.