Product Feeds for AI Shopping

Product feeds for AI shopping are the structured catalog data (files or API updates listing each product's title, price, availability, images, variants and policies) that merchants send to AI platforms so their products can appear in AI-generated shopping answers. They extend the feeds retailers have long sent to Google Shopping to newer destinations such as ChatGPT, Microsoft Copilot and Gemini, where an assistant picks a handful of products to show instead of a results page.

Feeds have become the main way into these answers. OpenAI says shopping in ChatGPT, live for US users, is powered by product feeds shared through the Agentic Commerce Protocol. Its commerce documentation recommends sending the full catalog once a day as a file and pushing price and stock changes during the day through an API; onboarding is open to approved partners, and Shopify and Etsy catalogs are connected automatically. In September 2026, ecommerce analyst Juozas Kaziukėnas reported that ChatGPT's shopping results now came entirely from merchant-submitted feeds rather than scraped pages and Google Shopping listings, so brands that send no feed have little chance of appearing. On Google, feeds go into Merchant Center and the Shopping Graph that powers AI Mode and Gemini. Microsoft says Microsoft Merchant Center feeds inform organic Copilot results, and Copilot Checkout runs through PayPal, Shopify and Stripe integrations.

A feed built for AI answers needs more than the classic Shopping minimum:

• Stable identifiers (item ID, GTIN, brand, MPN) and one row per purchasable variant, grouped so size and color options stay connected.
• Titles and descriptions that state the attributes people ask about, such as material, dimensions, compatibility, use case and who it is for.
• Current price, sale price and availability, updated as often as they change, not just in a nightly file.
• Clean images that show the actual variant, plus shipping times, return windows and ratings where the platform supports them.
• Optional question-and-answer and related-product data where available, such as Google's conversational attributes.

Consistency matters as much as completeness. The feed, the product page and its Product schema markup should agree on price, stock and specs. When they differ, assistants may show a stale price or skip the product, and platforms can remove items whose data does not match the landing page. The best setup uses one source of truth (a PIM or the commerce platform) that generates every platform's feed.

Feeds also connect discovery to transactions. In agentic commerce, the agent compares products from feed data and may hand off to checkout through ACP, Google's Universal Commerce Protocol or a merchant's own site. A product with incomplete feed data is not only less visible; it is harder for an agent to buy.

To measure feed performance, track whether your products appear in AI product cards for your category prompts, at what position and with which price and image. Promptwatch's ChatGPT shopping tracking records the product cards in ChatGPT answers for tracked prompts. Combine that with platform diagnostics, such as Merchant Center errors, and with AI-referred traffic and conversions in analytics. A product that never shows up for prompts it clearly fits usually has a feed gap: a missing attribute, a wrong category or an out-of-date price.