If you advertise on Google Shopping, or if your listings show up in Google's free results, you already maintain a product feed in Google Merchant Center. Title, price, availability, image, brand, GTIN: each product is described according to a strict specification that Google checks.
Few stores use that feed beyond advertising. It can also feed your site's internal search engine. This article explains what a feed contains, why it makes a good foundation for search, which data weighs most on result quality, and where it reaches its limits.
What a Merchant Center feed contains
A feed is a file, or a source connected to your CMS, that describes each product with standardized attributes. The ones that matter for search fall into four families.
Product identity
The id, title, description, brand, gtin and mpn attributes. Title and description are the two text fields the engine searches first. Brand and identifiers are used to find a specific product, by reference or by barcode.
Classification
The google_product_category and product_type attributes. The first follows Google's fixed taxonomy, the second reflects your own category tree. They feed category filters and help the engine understand which universe a query belongs to.
Choice attributes
The color, size, material, pattern, gender and age_group attributes. These become filtering facets and tell the variants of a single product apart.
Price and availability
The price, sale_price and availability attributes. They make it possible to sort, to filter by budget, and to keep an out-of-stock product from showing up first in the results.
One attribute deserves a mention of its own: item_group_id, which groups the variants of a single product. Without it, a t-shirt offered in twelve size and color combinations appears twelve times in the results. With it, the engine can show a single item and signal that variants exist.
Why it makes a good foundation for search
Three reasons add up.
A structured, validated source. Google rejects or flags products whose attributes are missing or malformed. That check, imposed for advertising, works in your favor: the feed is often cleaner than the raw CMS export.
A single source of truth. Prices, availability and titles are already kept up to date for Google Shopping. Plugging search into the same feed avoids maintaining a second catalog, with the risk of divergence: a product at €49 in search and €59 on its page.
Independence from the CMS. Whether the store runs on Shopify, PrestaShop, WooCommerce or Magento, the feed has the same format. That is what makes it possible to install a search engine with a simple script, without a separate module per platform.
From feed to engine: what happens
When an engine such as Google Vertex AI Search for Retail receives a feed, it does not just store it. It turns each product into a record in its catalog, indexes the text fields for keyword search, computes vector representations for search by meaning, and keeps the attributes for filters, sorting and facets.
Variants that share the same group identifier are attached to a primary product, which avoids duplicates. The engine can then display a single item with a "from" price when prices differ from one variant to another.
Vectail automates this path. You connect your Merchant Center account, enter your merchant ID, and the catalog is imported into Google Vertex AI Search for Retail, then resynchronized automatically at the frequency you choose. No export to schedule, no file to upload. An engine you integrate yourself, such as Elasticsearch, requires you to write that bridge by hand: it is one of the points covered in our Doofinder vs Elasticsearch comparison.
Feed quality sets the ceiling for results
An AI engine does not make up for poor data, it exposes it. Here are the feed defects that weigh most on search. The product page side is covered in our article on product data quality.
Titles that are too short or too generic
A title that contains no material, no cut and no color deprives the engine of everything your customers actually type.
Descriptions that are missing or copied from the manufacturer
The description field feeds search by meaning. An empty text, or the same text across a hundred products, erases the differences the engine is supposed to use.
An empty or inconsistent product_type
Without a clean tree, there is no reliable category filter, and the engine is less sure which universe a query belongs to.
Missing variant attributes
Without color and size filled in, there are no usable facets, and variants stay invisible to search and to filters.
Availability and prices out of step
A feed updated once a day while stock moves by the hour surfaces out-of-stock products, or displays an old price.
Where the Shopping feed reaches its limits
The feed is designed for product advertising, not for internal search. Four points are worth keeping in mind.
What is not in the feed does not exist for the engine. Products not submitted, rejected by Google or deliberately excluded from campaigns are missing from search unless you watch for it. Compare the number of products in your store with the number accepted in Merchant Center.
Titles written for Google Shopping. Some stores load titles with keywords or put the brand first for bidding. That does not block search, but a title a human can read is still preferable: it is the one displayed in your results.
No editorial content. The feed describes products, not your buying guides or your category pages. Surfacing that content in search goes beyond the scope of a product feed.
Attributes designed for Shopping. Some criteria that matter to your customers, such as compatibility or precise dimensions, have no standard field. They go through custom attributes, for example product_detail or the custom_label fields, and must be filled in explicitly if you want to filter on them.
Auditing your feed before plugging in an engine
Six quick checks, some of which already show up in the Diagnostics tab of Merchant Center.
1. Count. Compare the number of products in your store with the number accepted in Merchant Center.
2. Reread twenty titles picked at random: do they contain the product type, the brand and the attributes that matter?
3. Look for empty descriptions or ones duplicated from one product to another.
4. Check the product_type: a consistent tree, with no vocabulary variants for the same category.
5. Verify the variants. All the versions of a product must share the same item_group_id.
6. Compare the rhythms. How often is the feed updated, and how often does stock really change?
Is your feed ready? AI search installs in 2 minutes
Vectail connects to your Google Merchant Center, imports your catalog and turns on search powered by Google Vertex AI. One line of code to install, no plugin.
Start free - 14 days, no credit cardOur position
Vectail chose to build its onboarding on Merchant Center. It is a deliberate constraint: without a Merchant Center account, there is no automatic import. In exchange, the catalog comes from a feed already validated by Google, the same source serves Shopping and search, and the installation does not depend on any platform. The Merchant Center account itself is free.
Working on your feed is not one more project: it is the same one that improves your visibility on Google Shopping and the relevance of your internal search. If you only did one thing this week, reread twenty titles picked at random from your feed. It is the quickest check, and the one that reveals the most problems. The ones that then show up in your searches with no results are handled with our guide to zero-result pages.