Google Merchant Center Conversational Attributes, Business Agent, and AI Performance Insights: A Merchant Playbook

Merchant Center Is Now an AI Optimization Tool

For years, Google Merchant Center was the place you uploaded a feed so Shopping ads and free listings would work. In 2026 it quietly became something else: the main control panel for how your products show up in Google's AI shopping surfaces — AI Mode in Search, AI Overviews, the Gemini app, and branded chat experiences.

Three additions drive that shift:

  • Conversational attributes — six optional feed fields designed to help AI systems answer shopper questions about your products.
  • Business Agent — a branded AI sales associate that answers questions about your catalog inside Google's surfaces.
  • AI performance insights — reporting on how your products are discovered across AI Mode, AI Overviews, and Gemini, including share of voice against similar brands.

This post goes deeper than our Google AI Mode overview: what each feature does, what Google's documentation says, and how to prioritize.

The Six Conversational Attributes

Google announced conversational attributes in late May 2026 and has since expanded the documentation with formatting guidance and a TSV template. According to Google's Merchant Center Help documentation, they are optional, they won't affect the approval status of existing products, and they can be submitted through a supplemental data source (Google's recommended route), your primary data source, or the Merchant API.

Google names AI Mode, Gemini, and Business Agent as places this data helps, and notes it can also improve traditional search. Here is what each attribute is for.

1. question_and_answer

Product-specific FAQs as question-and-answer pairs. Google's example is a question about Bluetooth support paired with a direct answer.

This is the highest-leverage attribute for most stores. Shoppers in AI Mode ask questions, and this field lets you supply the answer in your own words instead of hoping the model infers it. Pull questions from real sources: support tickets, pre-sale chat logs, review text, and return reasons. If you already maintain on-page FAQs, see Product FAQs That AI Shopping Assistants Actually Cite for how to choose and phrase them.

2. document_link

URLs to related PDF documents such as manuals or assembly instructions. One or more links, comma-separated.

For technical or assembly-heavy products, this exposes authoritative detail that rarely fits in a description.

3. related_product

A group attribute linking an item to other products, with three sub-attributes: relationship_type, identifier_type, and identifier. Google's examples include a required part referenced by product ID and an accessory referenced by GTIN.

This is how you tell Google that a filter fits a specific purifier. Compatibility is exactly what conversational search handles poorly without explicit data.

4 and 5. item_group_title and variant_option

item_group_title gives a shared parent title to a product that comes in multiple variants (Google's example: "Organic Cotton Men's T-Shirt"). variant_option specifies what distinguishes each variant using name and value sub-attributes, such as shoe width and size. Both are used with item_group_id.

Together they let AI surfaces understand a product family as one item with options rather than a pile of near-duplicate listings. If your variant data is messy, start with Product Variant Data for AI Search.

6. popularity_rank

A number indicating how well a product performs relative to your other products, expressed as a percentage of total inventory (Google's example is 95.5). Higher means more popular within your own catalog.

This is a self-reported signal, so keep it honest and derive it from actual sales or order data on a regular schedule. A stale or inflated rank is worse than none.

One rule applies to all six: Google says not to repeat details already in description, product_highlight, or product_detail. Conversational attributes should add information, not restate it.

Business Agent: Your Branded AI Sales Associate

Google introduced Business Agent in January 2026 alongside the Universal Commerce Protocol. It lets shoppers chat with a branded agent that answers questions about your products inside Google's surfaces.

Eligibility

Per Google's announcement and coverage from Search Engine Roundtable at launch, a business needs to:

  • Be located in the U.S.
  • Have a verified Merchant Center account
  • Have at least 50 approved offers
  • Have a claimed brand profile

Setup

In Merchant Center, go to Marketing > Business Agent. Until you customize it, shoppers see a standard Google-branded version. Customization covers the agent's identity, welcome message, conversation starter prompts, colors, and support handoff contact details.

What it draws on

Business Agent uses your Merchant Center shopping data and your website, so thin product pages, outdated policies, or contradictions between site and feed will show up in its answers.

Google has said more capabilities are planned, including training the agent on your own business data (such as brand guidelines and size guides), customer insights, and agentic checkout within the chat. Treat those as roadmap items until they appear in your account.

How to prepare

  • Write your conversation starters around real buying questions, not slogans.
  • Make your shipping, returns, and warranty pages explicit and current — these are common questions and the agent will draw on your site. Our guide to shipping and return policy signals covers what to state.
  • Set up support handoff so questions the agent cannot answer reach a human.
  • Test it yourself with the questions customers actually ask, and fix the source data when answers are wrong.

AI Performance Insights: Measuring What You Fix

Google announced AI performance insights for Merchant Center in late May 2026. According to Google's help page and coverage at announcement, the reporting shows how products are discovered across AI Mode, AI Overviews, and the Gemini app, and includes a share of voice view that compares your brand with similar brands on shopping journeys that start from those surfaces.

Google said it would launch first in the U.S., Canada, Australia, India, and New Zealand, with its help page describing the feature as "coming soon." Availability may vary by account, so check your own Merchant Center rather than assuming you have it.

When it is available, use it as a feedback loop:

  1. Record a baseline before shipping conversational attributes.
  2. Change one category at a time so you can attribute movement.
  3. Compare share of voice over weeks, not days — AI surfaces fluctuate.
  4. Cross-check with off-Google measurement. Merchant Center only sees Google. ChatGPT, Perplexity, Claude, and Copilot need separate tracking, which is what our visibility tracking service covers.

Where UCP Checkout Fits

Conversational attributes and Business Agent are about being found and understood. Google's Universal Commerce Protocol is about being bought inside AI Mode and Gemini. Google's documentation describes UCP-powered checkout as available to select U.S. merchants, with merchants remaining the seller of record. Shopify has been enrolling eligible U.S. merchants through its agentic storefronts.

The order matters: checkout inside an AI surface does nothing if the AI never recommends your product. Fix the data layer first. For protocol background, see Agentic Commerce Protocols: ACP vs AP2.

A Practical Rollout Plan

  1. Audit core feed health. Disapprovals, missing GTINs, and price or availability mismatches undermine everything else.
  2. Create a supplemental data source for conversational attributes so you can iterate without touching your primary feed.
  3. Start with question_and_answer on top sellers, then add related_product, variant fields, document links, and popularity rank.
  4. Enable and customize Business Agent if eligible, and test it against real customer questions.
  5. Measure with AI performance insights where available, plus cross-engine tracking.

The Bottom Line

Google is telling merchants, through its own product decisions, that the feed is now the knowledge base for its AI shopping surfaces. The new attributes are optional, which means many competitors will leave them empty. That makes them an opening for stores willing to do the data work.

If you want help prioritizing, an AI visibility audit maps which products and attributes are holding you back, and our product optimization work fills the gaps across Google and the other AI engines.

Want to see how AI engines perceive your brand?

Get an Audit for $87