AI shopping answers do not read your website. They read your Google Merchant Center feed. If your GMC product data is incomplete or stale, AI Mode, Gemini, and other AI shopping surfaces recommend your competitors instead. Here is how the pipeline works and how to fix your feed.
Google Merchant Center (GMC) has become the main data source for AI shopping recommendations. When someone asks Google’s AI Mode for “a waterproof backpack under $80 that ships this week,” the answer is not built from your product pages. It is built from the Shopping Graph, Google’s index of more than 50 billion product listings, and that graph is fed primarily by Merchant Center. Your feed is now doing the job your product pages used to do. That changes what feed management means: it is no longer catalog housekeeping, it is visibility work.
What Is Google Merchant Center and Why Does It Matter for AI Search?
Google Merchant Center is the free platform where retailers upload structured product data: titles, descriptions, prices, availability, images, identifiers, and dozens of optional attributes. That data flows into Google Shopping, free listings, Shopping ads, and now every AI shopping surface Google runs. AI Mode in Search, shopping answers in the Gemini app, virtual try-on in Google Lens, and agentic checkout all pull from the same Shopping Graph that Merchant Center populates.
The Merchant Center feed is the canonical product source for Google’s AI answers. Schema.org Product markup on your pages acts as a supporting signal, and the two need to agree. When your on-page structured data says one price and your GMC feed says another, Google trusts neither with full confidence. The feed and the page should pull from the same source of truth in your catalog. Google documents the on-page side in its product structured data guidelines, and keeping both aligned is part of any serious technical setup, like the work we cover in our technical SEO services.
How Does GMC Feed Google’s AI Shopping Answers?
The pipeline has three layers. Your Merchant Center feed is the data layer. The Shopping Graph is the index layer, refreshing around 2 billion listings every hour. AI Mode and Gemini are the answer layer, where a language model reads the graph and assembles a recommendation in response to a conversational query. If your products are not in the feed, they are not in the graph, and the AI cannot recommend them no matter how good your website is.
One detail trips up a lot of stores: free listings must be enabled. Organic AI shopping answers draw from free Shopping listings, not paid campaigns. Plenty of retailers run Shopping ads for years without ever activating free listings, which leaves them invisible to organic AI recommendations. Check it under the programs section of your Merchant Center account before you touch anything else.
Which Merchant Center Attributes Do AI Systems Weight Most?
Attribute completeness is the biggest lever for Merchant Center AI visibility. Queries in AI Mode are on average three times longer than traditional searches, according to Google. “Show me a lightweight sunscreen that works under makeup” only matches products whose feeds actually contain weight, finish, and use-case data. A product with five filled attributes loses to a competitor with twenty-five filled attributes every time the question gets specific.
In January 2026, Google announced dozens of new GMC attributes built specifically for conversational commerce across AI Mode, Gemini, and Business Agent. Several carry an explicit note in Google’s documentation saying they are primarily intended for conversational experiences. The ones worth prioritizing: question and answer pairs (so the AI can answer buyer questions directly from your feed), popularity rank, document links for manuals and spec sheets, and related product references. Alongside those, the classic fields still carry the most weight: accurate GTIN, precise availability states, honest pricing, material, color, and size. A wrong or missing GTIN drops you out of the product clusters where AI answers compare retailers.
A weekly feed update is no longer enough. The Shopping Graph refreshes constantly, and AI answers prefer live data. If your feed says a product is in stock and it is not, agentic checkout fails, your reliability drops, and the AI shows you less often. Daily updates are the minimum. Real-time sync through the Merchant API is the target, and note that the older Content API for Shopping shuts down on August 18, 2026.
What Changed in Google Merchant Center in 2026?
Merchant Center now measures your AI visibility natively. In May 2026, Google announced AI performance insights inside GMC: share of voice benchmarks against similar retailers, a shopping funnel view split into discovery, evaluation, and purchase, the actual conversational terms shoppers use, and a report flagging which structured attributes your products are missing. The rollout starts in the United States, Canada, Australia, India, and New Zealand. For the first time, feed quality has a native scoreboard.
Two more changes matter. Google launched the Universal Commerce Protocol (UCP) with Shopify in January 2026, the layer that lets AI answers move from recommending a product to actually handling cart, checkout, and order tracking. And in May 2026, GMC added a beta option that scans your website with AI to create listings without a structured feed. Convenient for tiny stores, but a one-time scan cannot match the attribute depth of a maintained feed, and depth is exactly what conversational queries reward.
Your product feed now does the job your product pages used to do. Treat it like content, not like a compliance export.
How Do You Audit a GMC Feed for AI Visibility in Five Steps?
Check Diagnostics first. A disapproved product is removed from the Shopping Graph entirely, which means it vanishes from AI Mode, Gemini, and every other AI surface at once. Resolve disapprovals before optimizing anything, and build a weekly Diagnostics review into your routine.
Verify free listings are active. This single toggle decides whether organic AI shopping answers can include you at all. Paid campaigns do not substitute for it.
Score attribute completeness on your top products. Start with your best sellers and fill every recommended attribute, not just the required ones. Add the conversational fields Google introduced in January 2026, especially question and answer pairs.
Rewrite titles and descriptions in natural language. Describe what the product is, who it is for, and what it is made of, in plain sentences. AI answers are assembled from feeds that read like answers, the same principle we apply to written content in our guide on writing content for AI summaries and AI retrieval.
Align feed data with on-page structured data and increase update frequency. Price, availability, and ratings must match between GMC and your Schema.org markup, and updates should run at least daily. If you want the feed work connected to a broader AI visibility strategy, that is the core of our generative engine optimization services.
For Google’s AI surfaces, effectively yes. AI Mode and Gemini shopping answers are built from the Shopping Graph, and Merchant Center feeds are the primary way products enter that graph. On-page Schema.org Product markup supports and confirms the feed data, but it does not replace it. Without an active GMC feed with free listings enabled, your products are invisible to organic AI shopping recommendations on Google.
Indirectly, and increasingly through shared standards. The Universal Commerce Protocol that Google launched with Shopify in January 2026 is being adopted beyond Google’s own surfaces, and other AI shopping experiences rely on the same structured product data disciplines: complete attributes, accurate availability, and fresh pricing. Cleaning up your GMC feed usually improves the same catalog data that every other AI channel reads.
Daily at minimum, real-time where possible. The Shopping Graph refreshes about 2 billion listings per hour, and AI answers favor data they can trust right now. Stale availability is especially damaging because agentic checkout depends on it: a failed purchase attempt against an out-of-stock product hurts your reliability signals. Use the Merchant API for programmatic real-time updates, since the older Content API for Shopping shuts down on August 18, 2026.
Announced in May 2026, they are native GMC reports showing how your products perform across Google’s AI shopping experiences. They include share of voice benchmarks against similar retailers, funnel performance split into discovery, evaluation, and purchase stages, the conversational product terms shoppers actually use, and attribute gap flags for missing structured details like color, material, or style. The rollout begins in the US, Canada, Australia, India, and New Zealand.
Only as a starting point for very small catalogs. The beta scan Google added in May 2026 detects what is visible on your pages, which rarely matches the attribute depth of a maintained feed with dedicated fields for material, size, color, and category. Since conversational AI queries reward attribute depth, a scanned catalog will lose detailed comparisons to a properly structured one. Use the scan to get live fast, then migrate to a real feed.