Google just told merchants that product data quality determines whether they exist in AI commerce at all. That is a different sentence than anything they have said about Shopping before.
Google is adding a new category of product data fields to Merchant Center called “conversational attributes.” These are structured fields designed specifically for how AI retrieves products in natural language conversations, not keyword searches.
Google VP Courtney Rose confirmed during the 2026 Retail Ads Decoded session that Merchant Center product data now feeds seven surfaces: AI Mode, Gemini shopping, virtual try-on via Google Lens, Business Agent, brand profiles, free listings, and Shopping ads. One feed, seven surfaces. Conversational attributes are built for the AI-powered ones.
If you manage product feeds, this changes how you think about data architecture.
Conversational attributes are structured product data fields optimized for conversational AI retrieval. They go beyond standard Merchant Center fields like title, price, GTIN, and brand to include the kind of detail an AI needs to answer natural language shopping queries.
Traditional Shopping ads match keywords to product titles. AI parses structured attributes to answer queries like “What are the best affordable trail running shoes for wet conditions?”
To answer that, the AI needs to evaluate waterproofing specs, weight, terrain compatibility, and price as separate, queryable fields.
Google hasn’t published the full conversational attributes spec yet. But based on how AI Mode currently retrieves products, there are clear steps to prepare.
Audit your attribute depth. Count structured attributes per product in your feed. Under 15 means you’re likely invisible in AI Mode for anything beyond basic queries.
Add use-case and context fields. “Intended use,” “best for,” “season,” “compatible with” map directly to conversational queries. Include them in supplemental feeds or custom attributes.
Think in questions, not keywords. For each product, write down the five most likely natural language questions a shopper would ask. Check if your feed data could answer them. Gaps point to missing attributes.
Use specifics, not descriptors. “Waterproof” is a descriptor. “IPX4 rated, seam-sealed, tested to 10,000mm water column” is a specification. AI agents parse specifications.
Google has been telling merchants for years that product data quality matters for Shopping ads. Now Google is telling merchants that product data quality determines whether they exist in AI commerce at all.
The conversational attributes rollout is coming. The merchants who start enriching now will be ready. The ones who wait for the formal spec will be catching up.
Check where your products stand with a free AI readiness score at paz.ai. Enter any product URL, get a full breakdown of what AI agents see and what’s missing in 30 seconds.
Dor Shany – CEO of Paz.ai
Conversational attributes are structured product data fields designed for AI retrieval rather than keyword matching. Standard Merchant Center fields like title, price, GTIN, and brand were built for a system that matches search terms to product titles. Conversational attributes go deeper, covering use-case context, compatibility data, specification-level technical detail, and lifestyle fit fields that allow an AI agent to evaluate whether a product answers a shopper’s natural language query. The key difference is that AI Mode and Gemini Shopping do not retrieve products by matching keywords. They evaluate structured attributes to determine which products best answer a specific question, which means products without sufficient attribute depth simply do not appear in AI-powered results.
Based on how AI Mode currently retrieves products, fewer than 15 structured attributes per SKU is a strong signal that your products are likely invisible in AI Mode for anything beyond the most basic queries. This is not an official threshold from Google, but it reflects the practical reality that conversational queries require enough attribute data to evaluate product fit across multiple dimensions simultaneously. A shopper asking about trail running shoes for wet conditions needs the AI to evaluate waterproofing rating, terrain compatibility, weight, and price as separate fields. If those fields are not in your feed as discrete structured values, your products cannot be confidently included in the AI’s answer. Audit your current attribute count per product and prioritize enrichment for your highest-revenue SKUs first.
Google VP Courtney Rose confirmed at the 2026 Retail Ads Decoded session that one Merchant Center feed now powers seven surfaces: AI Mode, Gemini Shopping, virtual try-on via Google Lens, Business Agent, brand profiles, free listings, and Shopping ads. The AI-powered surfaces, specifically AI Mode, Gemini Shopping, and Business Agent, are the ones that require conversational attribute depth to surface your products effectively. Shopping ads and free listings continue to operate on keyword-to-title matching logic, though richer attribute data improves performance there as well. Virtual try-on and brand profiles have their own requirements around image quality and brand identity fields.
The most practical path for Shopify merchants is supplemental feeds submitted through Google Merchant Center. Supplemental feeds allow you to add custom attributes and additional structured fields to your existing product data without rebuilding your primary Shopify feed from scratch. Use-case fields like “intended use,” “best for,” “activity type,” and “compatible with” can be added this way, as can specification-level technical detail that is not captured in your standard Shopify product data. Feed management tools like Feedonomics, DataFeedWatch, and GoDataFeed can help you structure and manage supplemental feed submissions at scale. For a quick read on where your current product data stands with AI agents, paz.ai offers a free product URL audit that shows what AI agents see and what is missing.
A descriptor is a qualitative label: “waterproof,” “durable,” “comfortable,” “fast.” A specification is a measurable, verifiable value: “IPX4 rated, seam-sealed, tested to 10,000mm water column,” “1,000-denier Cordura nylon,” “memory foam with 3-inch profile and CertiPUR-US certification.” AI agents parse specifications because they can evaluate them against a shopper’s stated requirements. They cannot reliably evaluate descriptors because descriptors are subjective and do not map to specific shopper needs. When you are enriching your feed for conversational AI retrieval, the goal is to replace or supplement descriptors with specifications wherever your product has measurable attributes that could answer a specific question a shopper might ask.
