Google Merchant Center AI Insights Adds Conversational Search Terms, Intent & Attributes

Google has substantially enhanced its AI Performance Insights reporting within Google Merchant Center, unmasking the critical telemetry retailers need to understand how their products are discovered in generative AI search environments. The update introduces three new optimization dimensions: conversational search terms, structured shopping stage intents, and exact product attribute extractions.

Concurrently, Google confirmed that AI Performance Insights—which was previously restricted to US-based merchant accounts—is now actively available to retailers in Australia, Canada, India, and New Zealand. The expansion equips global merchants with the analytical tooling required to optimize their product catalogs for AI Mode and AI Overviews.

Traditional ecommerce search reporting has historically relied on keyword clusters—isolated terms like "running shoes" or "leather backpack". However, as consumers interact with multimodal assistants and generative search interfaces, user queries have shifted toward multi-sentence conversational questions containing nuanced situational constraints.

The upgraded AI Performance Insights console introduces three specialized analytics modules to decode this behavior:

1. 1.Top Conversational Terms: Highlights the specific contextual phrases and descriptive parameters shoppers include in natural-language prompts (such as "wide toe box running shoes for flat feet marathon training").
2. 2.AI Search Intents by Shopping Stage: Automatically segments conversational traffic across three distinct purchasing funnel stages:

• Discovery: Early exploratory queries where shoppers evaluate broad product categories and high-level concepts.
• Evaluation: Comparison queries where users compare competing models, evaluate technical specifications, or read aggregated reviews.
• Ready to Buy: High-intent, transactional queries focused on immediate price parity, warranty terms, and local stock availability.

1. 1.Popular Product Attributes: Pinpoints the exact structured attributes (such as heel-to-toe drop, fabric breathability, or battery cycle longevity) that Gemini highlights during recommendation synthesis—and identifies missing attributes from the merchant’s product feeds.

For digital retail leaders and catalog architects, the introduction of conversational attribute tracking resolves a long-standing black box in generative commerce. Historically, when an AI Overview recommended a specific SKU, merchants could not determine which product feed attributes triggered the inclusion.

With this release, Google provides direct algorithmic transparency. Retailers can now audit whether their structured data feeds are supplying the granular attributes Gemini requires to formulate confident recommendations. If a competitor is winning AI Overview citations for conversational queries, the scorecard explicitly highlights which missing attributes or specification gaps caused the omission.

To maximize product inclusion and commercial conversion across Google’s generative AI ecosystem, ecommerce teams should execute the following protocol:

• Enrich Feed Attributes Beyond Google Defaults: Do not limit product data feeds to basic mandatory fields (title, price, availability). Ingest deep technical specifications, materials, and dimensional attributes into supplemental feeds.
• Audit Stage-Specific Visibility: Review the Shopping Stage Scorecard to identify where conversion drop-offs occur. If a brand dominates Discovery but disappears during Evaluation, enrich comparison matrices and user review schema.
• Incorporate Conversational Language into Product Copy: Use high-velocity conversational terms identified in Merchant Center to naturally optimize product descriptions, FAQs, and buyer guides.

• About AI Performance Insights in Merchant Center — Google Merchant Center Help
• Google Expands Merchant Center AI Performance Insights — Search Engine Roundtable
• Google Helps Retailers Thrive With New UCP and AI Tools — Google Keyword Blog