Skip to content

Google Shopping Feed: One Feed for Ads and Your AI Assistant

· 7 min read · AskMerra

If you run Google Shopping ads, you already have a product feed: a file with every product, price and stock status, which Google reads on a schedule. The same file can power the AI assistant on your site. A well-kept feed works twice: your ads show the right price, and your assistant only recommends what you have in stock.

This guide explains what a feed contains, where to get one on each platform, which fields matter, and the mistakes that hurt both your ads and your assistant's answers.

What a product feed is

A feed is a file published at a web address, with one entry per product. The format Google Merchant Center expects is usually XML (RSS 2.0), with fields that start with g:. Some platforms also generate it as CSV. One product looks like this:

<rss version="2.0" xmlns:g="http://base.google.com/ns/1.0">
  <channel>
    <item>
      <g:id>SER-HA-30</g:id>
      <g:title>Hyaluronic acid moisturising serum for dry skin, 30 ml</g:title>
      <g:description>Light, fragrance-free serum for morning and evening...</g:description>
      <g:link>https://www.example.com/hydrating-serum-30-ml</g:link>
      <g:image_link>https://www.example.com/img/serum-30-ml.jpg</g:image_link>
      <g:price>24.00 EUR</g:price>
      <g:sale_price>19.00 EUR</g:sale_price>
      <g:availability>in_stock</g:availability>
      <g:brand>Example</g:brand>
      <g:product_type>Face care > Serums</g:product_type>
    </item>
  </channel>
</rss>

Google requires id, title, description, link, image_link, availability and price for all products, according to its product data specification. Other fields depend on the category and the country you sell in.

Where to get the feed on each platform

  • WooCommerce: the official Google for WooCommerce extension sends products to Merchant Center through Google's API. As WooCommerce's own documentation puts it, there is no feed file and no separate feed URL. You need a feed plugin that publishes the file at an address (usually shown in the plugin's settings) or a CSV export of your products.
  • Shopify: a feed app that publishes the file at an address, or the CSV export from the Products page. The AskMerra app for Shopify is in review for the Shopify App Store and will sync the catalog automatically. Until then, a feed is the simplest route.
  • PrestaShop: a Google Shopping feed module (several are on PrestaShop Addons) or a CSV export.
  • Magento 2: a feed extension, or the AskMerra Push API. The AskMerra module for Magento 2 is coming soon.
  • Any other platform (OpenCart or a custom build): if it can publish a Google feed or a CSV export, it works the same way.

One warning about exports: those made for re-importing into the same platform don't always include the link to the product page, and they use their own column names. Check the field mapping before you import. Without a link, a product card has nowhere to send the shopper.

The fields that matter

Field What Google asks What AskMerra does with it
id Required, up to 50 characters The key that updates are matched on. It must stay stable.
title Required, up to 150 characters The product name in answers and on the card.
description Required, up to 5,000 characters The basis for enrichment: use cases, specs, ingredients. HTML becomes text.
link Required The page the product card links to.
image_link Required The image on the card.
price Required, written like "24.00 EUR" The price, with its currency.
sale_price Optional The reduced price, shown next to the struck-through regular price.
availability Required: in_stock, out_of_stock, preorder, backorder Out-of-stock products are never recommended.
item_group_id For variants Groups sizes and colours of the same product.
brand, product_type, google_product_category Depends on the product The product's brand and categories.
color, size, material, gender, gtin Optional Become attributes the assistant can search on.

The last row is the one most stores underrate. When a shopper writes "red dress, size M, cotton", the assistant searches on attributes. If the feed has the colour, size and material, it finds the dress. If that information only exists in the photos, it can't know.

Common mistakes

1. Generic titles. "Serum 30 ml" says nothing. "Hyaluronic acid moisturising serum for dry skin, 30 ml" says what it is, who it is for and how much you get. Google can match it to more searches, and the assistant can recommend it more precisely.

2. Empty or very short descriptions. After import, AskMerra analyses each product to pull out what shoppers ask about: use cases, skin or hair types, ingredients, specs. A one-line description gives it nothing to work with.

3. Missing or wrong stock. If a product has no availability field at all (and no quantity), AskMerra treats it as available. preorder and backorder also count as available. Make the feed tell the truth about stock, because the assistant relies on it.

4. Expired sales left in the feed. Google can also read the sale period (sale_price_effective_date). AskMerra reads sale_price exactly as it appears in the feed and does not read the period. When a promotion ends, the sale price has to leave the feed.

5. IDs that change. If products get new IDs after a migration or a plugin switch, every system sees them as new products. In AskMerra, the old ones are deactivated and the "new" ones are processed again, which uses AI.

6. A feed filtered for one campaign. By default, AskMerra deactivates products that disappear from the latest feed (it never deletes them). If you removed a category from the feed for your ads, the assistant stops recommending it too. Give the assistant a feed with the whole catalog, or switch off the option that deactivates missing products.

7. Prices without VAT. Some feeds are built for price comparison sites that add VAT themselves. For the assistant, use the feed with the price the shopper actually pays.

The same feed for your AI assistant

In AskMerra, you add the feed in the dashboard on the Catalog page, as a Product feed URL source. Paste the address, and AskMerra detects the format (Google Shopping XML, CSV or JSON) and suggests how your columns map to its fields. You review the mapping, then import. Rows with errors (missing id or name, invalid price) are skipped and listed in the import report, and the rest of the catalog comes in normally.

A few useful details:

  • The feed has to be at a public address, over http or https.
  • Descriptive Google fields such as color, size and material are suggested as attributes automatically. You can map other columns as attributes too, for example attribute:skin_type.
  • If you sell in several languages, add one feed per language or use a language column. The assistant answers in the shopper's language either way, even if your catalog exists in one language only.
  • After import, each new or changed product is enriched by AI and indexed for semantic search. You can import before you choose a plan; enrichment starts once you subscribe.

For the technical details, see the catalog sources docs and supported platforms.

How fresh it stays

The slowest link sets the pace.

  • AskMerra re-reads your feed automatically, somewhere between every hour and every six hours depending on your plan. The pricing page shows the exact interval for each plan.
  • Your platform regenerates the file on its own schedule. If it rebuilds the feed once a day, a price change can take up to a day to reach the feed, however often AskMerra reads it.
  • Price and stock changes apply as soon as they are imported and do not trigger a new enrichment. Unchanged products cost nothing to re-import.
  • Real-time stock: if you sell low-stock items that can sell out within hours, send changes through the Push API, which updates the catalog immediately.

Whatever the source, product cards always show the price and stock from the latest sync, never from the AI model's memory. We explain why that matters for accuracy in accurate AI answers grounded in your catalog.

Checklist

  • The feed is published at a public address and updates at least once a day.
  • Every product has an id that never changes.
  • Titles say what the product is, who it is for and which variant it is.
  • Descriptions include what a shopper would ask: material, dimensions, ingredients, compatibility.
  • availability reflects real stock.
  • sale_price appears only while the sale runs.
  • Prices are what the shopper pays, including VAT.
  • The feed for the assistant contains the whole catalog.
  • Colour, size and material sit in their own fields, not only in the title.

Connect your feed

Create your account, add your feed address on the Catalog page and check the field mapping. If you are installing the assistant for the first time, the guide to adding an AI shopping assistant to Shopify or WooCommerce covers the widget too. Large feed or several languages? Book a demo and we will look at your feed together.

Put AskMerra to work on your store

Connect your catalog, start your 14-day free trial and test the assistant in the playground before you go live.

Start free trial

Keep reading