How it works
It works like a product feed for Google Shopping or marketplaces: a structured export of the catalog, refreshed on a regular schedule.
What changes is the use. An AI assistant has to answer questions, and a title and a price aren't enough for that. So the feed needs usable attributes (dimensions, materials, compatibility, use cases), clear descriptions written in natural language, current price and availability for each variant, and a link to the right product page.
The more precisely the feed describes each product, the better the assistant can match it to a shopper's question.
Why it matters in e-commerce
Consumer AI assistants, ChatGPT included, now show products in their answers. Without reliable data on a brand, they leave it out or show the wrong price and stock.
The feed is how a brand controls what these assistants know about its products. It's also a direct GEO lever.
The stakes are already measurable. In the 2026 barometer iAdvize ran with Ifop (in French), 62% of AI users turn to it to learn about a brand or product, compare offers and get purchase recommendations. What the assistant answers depends on what the feed gave it.
In its webinar with Lengow and JVWEB on Google AI Overviews and AI agents (in French), iAdvize reports the same thing from its customers and the market: what AI Overviews, AI Mode and agents have in common is data. Retailers who had already started cleaning up their product data feel equipped, while the others are starting in a rush.
Further reading: Agentic Commerce Is Here, and What To Do About It
Real-life example
A skincare brand's feed only holds each product's name, price and image. A consumer asks ChatGPT for a fragrance-free moisturizer for sensitive skin. Nothing in the feed says "fragrance-free" or "sensitive skin", so the brand's creams don't come up. Once the feed includes those attributes, they can appear in that kind of answer.
Common confusion
Many brands assume their Google Shopping feed is enough. That feed is built for ads: title, price, image, category. An AI assistant has to answer questions, so it needs more detailed attributes and descriptions. Both feeds come from the same catalog, at different levels of detail.