Aizii Learn · AI Search

[← Back to Learn](https://aizii.ai/learn)

# How to use your product feed for AI search

A product feed is a file that lists your products, one row per product, with the same facts on every row: an id, a title, a description, a link, an image link, a price, and a stock status. AI search uses that file as a second source for your catalog, beside your pages. When the feed and the pages agree, an AI assistant can check a price or a stock status in the feed and cite the page. When they disagree, the assistant has two answers and trusts neither. This guide covers how to build a feed that AI search can use: the fields, the format, the variants, and the routine that keeps it current. It does not promise that a feed gets your products picked, or that a feed is read on any fixed schedule. No vendor publishes a fixed pickup time for a feed, and this guide does not invent one.

A feed is a developer-led job with one marketer check: the developer owns the file, the format, and the upload or fetch, and the marketer owns the titles, the descriptions, and which products belong in it. Agree the field list first, then build the file, before you submit anything.

## Where it goes wrong

Feeds fail quietly. The file uploads, the account shows no error, and the products that matter are still missing, wrong, or stale.

-   The feed is built from a different source than the pages. The feed comes from a spreadsheet or an old export, the pages come from the store database, and the two drift apart. The price in the feed is last week's price, and the page shows this week's price.
-   A required field is missing. A product has no id, no image link, no price, or no stock status, and the row cannot be used. Google's product data specification lists id, title, description, link, image link, availability, and price as required for every product, with brand and a product identifier such as a GTIN or an MPN where they apply. A row missing one of these is a row that cannot serve.
-   The title is a code, not a product. A title that reads only as an internal code or a style name gives the feed reader no product type, no feature, and no words to match against a shopper's question.
-   The variants are one row. A product in six colors and five sizes appears as a single row with one price and one stock status. A shopper who asks for one combination cannot be matched, because the feed never says which combinations exist or which are in stock.
-   The link points to the wrong address. A feed link goes to a preview domain, an old address, or a page that redirects, while the live product page sits somewhere else. A reader that follows the link finds a dead end or a different product.
-   The stock status is stale. A product sells out, the page says out of stock, and the feed still says in stock. The assistant recommends a product the shopper cannot buy, and the shopper learns not to trust the next recommendation either.
-   The file format is wrong. A feed saved in an unsupported format, or with the wrong delimiter, cannot be read at all. Google Merchant Center accepts tab-delimited text files and XML files, and files can be delivered by upload or by a hosted link that Merchant Center fetches on a schedule.

## The steps

**Step 1. Agree the field list before you build the file.** The marketer and the developer list the fields every row will carry: id, title, description, link, image link, availability, price, brand, condition where it applies, and a product identifier such as a GTIN or an MPN where the product has one. Add the fields your catalog needs on top: color, size, gender, and age group for apparel, and an item group id where a product has variants. Check the list against the Google product data specification in the Sources section. Do not add a field you cannot fill from the store database, and do not leave a required field blank.

**Step 2. Generate the feed from the same database as the pages.** The developer builds the feed from the same source that renders the product pages, so a price has one source and one value. Do not build the feed from a separate spreadsheet that someone updates by hand, and do not copy prices into the feed from the page. A feed built from the same database cannot drift from the pages unless the database itself is wrong, and then the fix belongs in the database, not in the feed.

**Step 3. Give every variant its own row.** The developer submits each size and color combination as its own item, with its own id, its own price, and its own availability, and groups the combinations of one style under one item group id. Google's Merchant Center guidance for free listings asks for this split for apparel, and the same shape works for any product with variants. A product with thirty combinations needs thirty rows, not one row that pretends to cover them all. Keep the item group id stable when you update the data, so the combinations stay grouped.

**Step 4. Write titles and descriptions the way shoppers ask.** The marketer writes each title to name the product, its key feature, and its color or size where that applies, in plain words, and each description to say what the product is, what it is made of, and who it is for. Keep the title the same in the feed, on the page, and in the markup. A title that changes between the feed and the page teaches the reader that neither version can be trusted. Do not put promotional words in the title, and do not put a price in the title.

**Step 5. Choose the format and the delivery method.** The developer picks the format that fits the catalog: a tab-delimited text file for a smaller catalog that a spreadsheet can hold, or an XML file for a larger or more complex catalog. Google Merchant Center accepts tab-delimited text files and XML files, and a single product file can be up to 4 GB. Deliver the file by upload for a one-off check, or by a hosted link that Merchant Center fetches on a schedule for a catalog that changes often. Check the Merchant Center help page for creating a product file, linked in the Sources section, before you pick a delivery method, and confirm the file extension matches the format you chose.

**Step 6. Check the feed against the pages before you submit it.** The marketer picks a sample of products, including one with variants and one that recently changed price, and the developer compares the feed row against the live page: the same title, the same price, the same currency, the same stock status, and the same link. Open each feed link in a browser and confirm it loads the right product page. Open each image link and confirm the image loads. Fix any row that disagrees with its page before you submit the file, not after.

**Step 7. Build the refresh into the catalog routine.** The developer sets the feed to regenerate when the catalog changes, and the marketer checks the feed when products are added, removed, or repriced. A removed product should leave the feed the same day it leaves the catalog. A price change should reach the feed the same day it reaches the page. A new product should appear in the feed when it is published, not weeks later. If the catalog changes several times a day, check the vendor docs for the API option for pushing product data instead of waiting on a file fetch.

## The machine-readable piece

This is a tab-delimited feed row for one variant of a hypothetical product, with the header row above it. The columns are the attribute names from the Google product data specification. Replace the placeholders in square brackets with your own values. The availability value uses the feed form, in\_stock, which is the value the specification asks for in a feed, not the schema.org form used in page markup. The price carries the number and the currency code, and the link is the full address of the product page. Each variant needs its own row, with its own id and its own availability, and all variants of one style share one item group id.

```
id	title	description	link	image_link	availability	price	brand	condition	item_group_id	color	size
[your SKU]	Copper saucepan, 20 cm, stainless steel interior	A copper saucepan with a stainless steel interior, a 20 cm diameter, and a stay-cool handle.	https://[yourdomain.com]/products/copper-saucepan-20cm	https://[yourdomain.com]/images/copper-saucepan-20cm.jpg	in_stock	89.00 USD	[Your brand name]	new	[your style id]	copper	20 cm
```

Check the file before you submit it: confirm the header row matches the attribute names above, and confirm every row has a value in each required column. Submit the file in Merchant Center, and fix any row the account flags.

## What it looks like when it works

On the page, nothing looks different to a human reader. A feed is not a page shoppers see. The change is under the surface: every product in the catalog has one row, every variant has its own row with its own price and stock status, every link loads the right product page, every image link loads, and the price and stock status in the feed match the page exactly. When a price changes, the feed changes the same day. When a product sells out, the feed says out of stock the same day. When a product is removed, it leaves the feed the same day.

Say you run a hypothetical kitchenware store selling saucepans, knives, and baking trays, with each saucepan in three sizes. The marketer writes each title to name the product, the material, and the size, and the developer builds the feed from the same database that renders the product pages. Each size of each saucepan is its own row, grouped by one item group id per style, with its own price and stock status. A shopper asks an assistant for a copper saucepan in 20 cm under a set price. The assistant can check the feed, confirm that size is in stock at that price, and cite your product page instead of a competitor's listing whose feed still shows last week's price.

### The fundamentals

These mechanics are shared across every store and covered in full in their own guides:

-   •The llms.txt format and a complete sample, see our [guide to writing a good llms.txt file](https://aizii.ai/learn/ai-search/how-to-write-a-good-llms-txt-file)
-   •The markup AI systems read to verify facts, see our [product schema guide](https://aizii.ai/learn/ai-search/how-to-add-product-schema-markup-for-ai-search)
-   •The complete crawler rule reference, see our [robots.txt guide for AI search](https://aizii.ai/learn/ai-search/how-to-write-robots-txt-for-ai-search)

## What this guide does not cover

This guide covers building and using a product feed for AI search: the fields, the format, the variants, and the refresh routine. It does not cover the full technical audit that checks every crawler and every page, the sitemap strategy for a large catalog, the transaction layer that lets an agent complete a purchase, or paid advertising. It does not cover the Merchant Center account setup in full, and it does not cover feeds for channels other than Google Merchant Center. Those are separate jobs, and this guide is the map for the feed, not the whole trip.

## Check your own store now

Run the free scan. One click, no signup, and it shows whether AI crawlers can read your store today: aizii.ai/aeo-scan

[Run the free scan](https://aizii.ai/aeo-audit)

## Sources

-   Google Merchant Center Help, Free listings for products (specific): [https://support.google.com/merchants/answer/9199328?visit\_id=639234295611713744-2772141442&rd=1](https://support.google.com/merchants/answer/9199328?visit_id=639234295611713744-2772141442&rd=1)
-   Google Merchant Center Help, Create a product file for Merchant Center (specific): [https://support.google.com/merchants/answer/12631822?sjid=8544664325339344309-EU&visit\_id=639275790238767947-1552424986&rd=1](https://support.google.com/merchants/answer/12631822?sjid=8544664325339344309-EU&visit_id=639275790238767947-1552424986&rd=1)
-   Google Search Central, Product structured data (specific): [https://developers.google.com/search/docs/appearance/structured-data/product](https://developers.google.com/search/docs/appearance/structured-data/product)
-   Google Search Central, Optimizing your website for generative AI features on Google Search (specific): [https://developers.google.com/search/docs/fundamentals/ai-optimization-guide](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide)
-   Schema.org, Product type (specific): [https://schema.org/Product](https://schema.org/Product)

Note: links marked specific point to the page that supports the claim in this guide. Links marked general docs point to the vendor's general docs section for that topic, not to a page that proves every step. Check the vendor docs for the current path before relying on a menu name or click path in this guide.
