The queries hitting Google's AI Mode are three times longer than traditional searches. One in six arrive via voice or image input, not text. If your Merchant Center feed is built around short keyword phrases, you're mismatched to where shopping discovery is now moving.
Google announced two new Merchant Center features at Marketing Live last week that make this problem visible and partly fixable: Conversational Attributes and AI Performance Insights.
What conversational attributes actually are
You can now add new structured fields to your Merchant Center product data. Google's AI uses these fields to answer natural-language shopping queries across AI Mode, the Gemini app, and AI Overviews.
The attributes include product_faq (common questions and answers about the product), product_use_cases (specific scenarios where the product fits, like "compact for carry-on" or "safe for toddlers"), product_substitutes (what Google should recommend when the item is out of stock), and native_commerce (a boolean that marks whether the product is eligible for agentic purchase completion inside AI Mode, with no site visit required from the customer).
These fields are submitted as a supplemental data source. Adding them won't affect your primary feed or product approval status. They're optional. Google's AI will still index products without them. But if a competing product has four detailed FAQ entries and yours has a generic 12-word description, you can predict which one surfaces when someone asks: "help me find a water bottle that fits a backpack side pocket and doesn't leak."
Why the existing feed doesn't hold up
Most Merchant Center feeds were built for keyword matching. Titles front-load the primary search term. Descriptions restate the title in slightly different word order. Categories are set once and rarely revisited.
That worked when users searched "blue water bottle 32oz BPA free" and Google matched those words to your title. AI Mode doesn't do that. It interprets the intent behind a longer, more specific question and synthesizes an answer from product data. The match is now question-to-attribute-coverage, not keyword-to-keyword.
A product with zero conversational attributes is harder for an AI system to confidently recommend. You get deprioritized not because your product is wrong for the query, but because the AI can't extract enough data to say it's right. That distinction matters more than it looks. Your product could be the best answer and still not appear.
The AI Performance Insights dashboard
Also launched at GML: AI Performance Insights, a new reporting section inside Merchant Center. It shows how your products are being discovered across AI Mode, the Gemini app, and AI Overviews.
Performance breaks down by four dimensions: journey stage (discovery, evaluation, or purchase intent), product terms (the phrases AI systems use when surfacing your products), structured attribute coverage (which feed fields are actually driving impressions), and catalog gaps (where your product coverage falls short of query patterns Google is seeing).
This is visibility you didn't have before. Most accounts had no idea whether their products were showing up in AI-mediated search at all. Now you'll be able to see it directly in Merchant Center. It's rolling out in the US, Australia, Canada, India, and New Zealand over the coming months.
What to do before the gap widens
Start with your top 20 SKUs by revenue. For each one, write four to six product_faq entries based on questions your support team actually gets. Add two product_use_cases covering the specific contexts where the product fits customers' real situations. Submit as a supplemental feed.
This isn't a feed rebuild. It's an addition. The goal is to give Google's AI the specific language it needs to match your products to conversational queries they'd otherwise miss.
When AI Performance Insights reaches your account, check journey-stage breakdowns first. Products that appear at evaluation but disappear at discovery almost always have thin attribute coverage at the awareness end. That gap costs more revenue than it shows on a dashboard.
Brands winning in AI-mediated shopping have product data rich enough for an AI to confidently answer a specific question about a specific product. That's a content problem dressed as a feed problem. The brands that treat their product data as marketing copy rather than a logistics spreadsheet are going to pull away from those that don't.
Google hasn't said when AI Performance Insights will reach all markets. Use the gap to get the data right before the visibility tool arrives.
If you want a clear read on how your account is positioned before AI Mode becomes the primary discovery surface, the free audit at Gromerce shows what's exposed in about three minutes.
Your product feed is the source of truth for AI shopping. Treat it like one.
Sources: Search Engine Land, Google Merchant Center Help, Google Blog, May 2026
