How to Use Merchant Center Conversational Attributes

Google announced on September 16, 2026 that AI performance insights are available in Merchant Center for eligible businesses in Australia, Canada, India, New Zealand, and the United States. The report covers English-language conversational shopping queries in AI Mode and AI Overviews, including shopper terms, intent, and attributes.

For ecommerce teams, the right response is not to stuff more copy into the feed. Use the report to find a real information gap, verify the answer against a reliable source, place it in the correct field, and test a controlled batch.

Quick answer

• Fix standard feed facts before adding optional conversational fields.
• Treat AI insights as a research queue, not a ranking score.
• Add only answers supported by current product evidence.
• Keep evidence, normalized facts, and channel presentation separate.
• Test a small group and record what changed.

1. Check core product data first

Google describes conversational attributes as optional fields that complement the primary Merchant Center product data specification. They do not replace the basics.

Audit one representative product from each important category:

• Product and variant titles identify the exact item.
• Price, currency and availability match the landing page.
• Pack quantity and included contents are explicit.
• Size, material, dimensions and compatibility use consistent values.
• Images show the correct variant.
• Shipping and return information is current.
• Identifiers map to the correct sellable item.

If the answer belongs in a standard attribute such as material, size, multipack, product detail or description, fix that source first. Google specifically says merchants do not need to repeat information in conversational attributes when it is already provided through descriptions, product highlights or product-detail fields.

The purpose is a richer product record, not a parallel feed that ages differently.

2. Turn AI insights into an evidence queue

Google's AI performance insights documentation says the report can show conversational search terms, shopper intent, and popular attributes. Availability currently applies to English-language queries for eligible Merchant Center accounts in five countries.

The report is useful for prioritization, but it is not a complete explanation of ranking or revenue. Record a baseline before changing the catalog.

For every proposed update, capture:

1. Market and date range.
2. Product or category.
3. Reported term, intent, or attribute.
4. Current feed and landing-page value.
5. Customer question being answered.
6. Evidence source and approval date.
7. Field to update.
8. Owner and review date.

Work in small batches. When many fields, descriptions and products change at once, neither a visibility gain nor a new disapproval has a clean explanation.

3. Choose the field that matches the job

Google's conversational-attributes guide lists six optional attribute groups: question and answer, document link, related product, item-group title, variant option, and popularity rank.

Use a field only when its job matches the information.

Google's September 16 retailer announcement encouraged merchants to keep feeds accurate and add conversational detail. Independent reporting from Search Engine Roundtable also noted the wider AI-insights availability and the emphasis on conversational attributes.

Neither source says more fields automatically guarantee placement.

4. Separate evidence, facts, and channel copy

A durable catalog uses three layers.

Evidence includes supplier specifications, manuals, measurements, certificates, test records, and dated approvals.

Normalized facts turn that evidence into consistent values: one unit system, one controlled material name, one pack quantity, and one verified compatibility statement.

Channel presentation sends appropriate facts to the storefront, Merchant Center, marketplaces, support tools, and AI-shopping channels.

Good example:

• The exact product manual confirms that an accessory is included.
• The normalized record stores `included_accessory: true` and the accessory identifier.
• Merchant Center receives a concise question-and-answer entry.

Bad example:

• A similar model includes the accessory.
• Marketing assumes this version does too.
• The assumption is copied into the feed and repeated by several channels.

The second workflow scales beautifully right up to the refund.

For broader channel governance, use our Shopify agentic storefront access audit to record which destinations can receive catalog data and who owns each decision.

5. Test a controlled update

Start with five to twenty products from one category rather than the entire catalog.

For each product:

• Confirm the evidence and current standard attributes.
• Add only the conversational field that answers a documented question.
• Submit through the recommended supplemental source, primary source, or Merchant API workflow used by your team.
• Confirm the processed product contains the intended value.
• Recheck the landing page for consistency.
• Record product issues, impressions, clicks, AI-insight changes, and support questions.

Allow enough time for processing and observation. Do not interpret a single query or a short-lived dashboard movement as proof.

If an answer changes with price, inventory, promotion, or date, it probably belongs in a maintained standard field rather than static conversational copy.

Final checklist

• [ ] Confirm account, country and language eligibility.
• [ ] Export a baseline of current AI insights.
• [ ] Fix title, variant, price, availability and identifiers first.
• [ ] Check whether the answer already exists in a standard field.
• [ ] Link every new statement to current evidence.
• [ ] Choose the conversational attribute that matches the job.
• [ ] Avoid duplicating price, availability or promotional copy.
• [ ] Assign a field owner and review date.
• [ ] Test a small product group.
• [ ] Verify processed feed data and landing-page consistency.
• [ ] Measure without promising rankings or AI placement.

FAQ

Do conversational attributes affect product approval?

Google says adding these optional attributes does not affect the approval status of existing products. Standard product requirements and shopping policies still apply.

Should every product have question-and-answer entries?

No. Add them where a real, product-specific question has a verified answer that is not already supplied through the appropriate standard fields.

Do conversational attributes guarantee visibility in AI Mode?

No. They can provide richer context, but Google controls eligibility and display. Complete data does not guarantee inclusion or ranking.

Can AI performance share of voice be treated as sales attribution?

No. Use it as a discovery and diagnostic signal. Evaluate sales, traffic, product issues, and customer behavior with their appropriate reports.

Start with one category and one evidence-backed question. A smaller catalog update that the team can verify, explain, and maintain is more useful than a heroic bulk edit whose main source is enthusiasm.

References:

• Google: September 16 retailer update
• Google Merchant Center: AI performance insights
• Google Merchant Center: Conversational attributes
• Search Engine Roundtable: Merchant Center AI updates