# Structured Data for AI Shopping: The Fields Answer Engines Actually Read

Source: https://aicommercewatch.com/blog/2026-11-14-structured-data-ai-shopping-fields.html

November 14, 2026

*A shopper asks ChatGPT for a product, the answer quotes a price, and the price is wrong. The store owner checks the product page and the price is right. The answer engine never read the product page. It read the Merchant Center feed, and the feed still had last week's price.*

## The feed is the product page now

ChatGPT shopping answers pull from Google Merchant Center feeds, not from product pages. That is the finding behind the [product feed post](2026-11-13-product-feed-ai-shopping.md) from yesterday: the carousel matches the feed, and the page is a secondary source. The implication is direct. The feed is no longer a distribution artifact for Google Shopping. It is the product record that answer engines read, and every field in it is a fact that can be quoted back to a customer. A store that treats the feed as a fire-and-forget export is letting a stale snapshot speak for its catalog.

The same shift is happening across answer engines. The [citations post](2026-11-11-ai-search-citations-answer-engines-product-pages.md) showed that product pages get quoted when they carry the right structured data, but the feed is the primary source for shopping surfaces. The page and the feed have to agree, and when they disagree, the feed wins.

## The fields answer engines actually read

Schema.org Product markup on the page carries name, description, price, priceCurrency, availability, image, brand, and the identifiers GTIN and MPN. The Merchant Center feed carries the same facts in its own attributes: id, title, description, price, availability, image link, and GTIN. The fields that decide whether a product appears at all are the identifiers and the availability. A product with no GTIN is harder to match to a shopping query. A product whose availability says in stock when it is not will be recommended, clicked, and then refused at checkout. A product whose image link is broken will not appear in a carousel, because there is nothing to show.

Reviews and ratings ride along too, when they exist. AggregateRating on the page and the review feed in Merchant Center both feed the stars that appear next to a product. A product with no rating data is not penalized. It is simply less convincing than the one next to it that has one.

## Where the errors creep in

The errors are mundane and they are all stale data. The price changed on the page and the feed still has the old one. The product sold out and the availability still says in stock. The image link points at a file that was renamed. The GTIN is missing because the supplier never provided it. Each one is a small failure on its own, and together they are a catalog that answer engines quote wrong. The [feed errors post](2026-11-10-ai-shopping-assistants-product-feed-errors.md) documented the same pattern from the assistant side: the assistant is only as accurate as the feed it reads.

## The audit: feed vs page, line by line

Once a week, pull the feed and compare it against the live pages: price, availability, image link, GTIN. Check the products that changed that week first, then a sample of the rest. After every price change, stock change, or product launch, check the affected products the same day. The audit takes an hour and it catches the errors that become customer complaints. A feed that matches the store is a catalog that gets quoted right. A feed that does not is a catalog that argues with its own customers.

## Frequently asked questions

### Do I need both a Merchant Center feed and schema.org markup?

Both, but they serve different readers. The Merchant Center feed is what shopping surfaces read, including ChatGPT shopping and Google Shopping. Schema.org Product markup on the page is what search engines and answer engines read. If the two disagree on price or availability, the feed usually wins, and the disagreement is what gets you quoted wrong.

### Which feed fields matter most for AI shopping?

The feed fields that decide whether a product appears are id, title, description, price, availability, image link, and GTIN. A product with no GTIN, a stale price, or an availability value that says in stock when it is not will either not appear or appear wrong.

### How often should the feed be checked against the store?

Weekly, at minimum, and after every price change, stock change, or product launch. The feed is a snapshot, and it goes stale the moment the store changes. A product that sold out on Monday and still shows as in stock on Friday is a customer service problem by the weekend.

### Can a feed price mismatch get my product quoted wrong?

Yes, and it is the most common way products get quoted with the wrong price. The feed price and the page price have to match, including the currency. A mismatch means the answer engine quotes one price and the checkout charges another, and the customer has the screenshot.