Google Conversational Attributes

AI-driven shopping is changing how buyers find and evaluate products — through AI Mode in Search, conversational agents, and other AI-powered surfaces. To help these systems represent your products accurately, Google Merchant Center supports a set of conversational attributes.

These attributes let you submit richer, more nuanced product information than a standard feed allows: answers to the questions buyers actually ask, links to related products and accessories, and detailed variant options such as size, color, and fit. Feeding this context to Google helps AI systems and conversational agents understand your products the way a well-briefed salesperson would, so buyers find the right product faster — whether they're browsing traditional search results or asking an AI assistant a question.

This feature is built for Productsup clients who already export a product feed to Google Merchant Center and want to enrich that feed with conversational depth, without touching or risking the integrity of their core product data. Because this export complements your existing Google Merchant Center feed rather than replacing it, it's available at no additional cost.

Available conversational attributes include:

Google recommends submitting conversational data as a supplemental data source, layered on top of your existing primary feed, rather than folding it into the primary feed itself. Keeping these attributes in a separate feed also makes the setup easier to manage and reduces the risk of formatting issues affecting your primary feed — particularly for complex attributes such as question_and_answer, where quotation marks can be misinterpreted during import.

Before setting up the export in Productsup, prepare a supplemental data source in Google Merchant Center:

See Google's guide to creating a supplemental data source for full details, and the conversational attributes reference for how each attribute works and how to format its values.

The numbered fields on structured attributes also let you use rules to decide which information populates each field, depending on the data available for each product.

The id field is required for this export. Map exactly the same product ID that your primary feed sends to Google Merchant Center. Google matches conversational data to existing products by ID — if the identifiers don't align, your enrichment data doesn't attach to the right products and the update has no effect.

question_and_answer provides product-specific questions and answers. This export supports up to 30 pairs per product, with separate fields for each pair:

Map the raw question and the raw answer into the corresponding fields — for example:

There's no need to add quotation marks or build Google's final syntax manually — the platform applies the required formatting during export.

document_link provides links to product-related documents, such as manuals, user guides, or assembly instructions. This export supports up to 5 links per product:

Map each document URL into the corresponding field — for example, document_link[1] → https://example.com/product-manual.pdf. The platform builds the required output structure during export.

item_group_title provides a shared title for products that belong to the same variant group. Describe the general product without variant-specific details such as size or color.

item_group_title requires an item_group_id on the product. Make sure the corresponding item group ID is available and mapped.

popularity_rank indicates a product's relative popularity, expressed as a value between 0.0 and 100.0.

Map your calculated value directly into popularity_rank — no additional formatting is required. Define the ranking logic based on your own business data, and apply a consistent methodology across your assortment so values stay comparable.

related_product links the current product to related products — cross-sells, accessories, or required parts. This export supports up to 30 related products, each defined by three fields:

identifier_type accepts:

relationship_type accepts:

For example:

Use the next numbered fields for additional related products, up to 30, each with its own complete set of values. Map only the raw values — the platform combines them into the format Google requires during export.

variant_option captures the specific information that distinguishes one product variant from another within the same item group. Which attributes qualify depends on your product data and how the products sharing an item_group_id differ — for clothing, that might be size, color, material, or fit.

This export supports up to 30 variant options, each with a separate name and value field:

Map only the raw name and value — the platform generates Google's required variant_option structure during export.

Suppose products in the same item group vary by size and color.

This produces, for example:

Continue the same pattern for additional variant options.

Rules are also useful for excluding attributes that don't actually distinguish the variants. For example, if every product in an item group shares the same size, size isn't a distinguishing attribute for that group — create a rule to leave the corresponding variant option empty in that case.

The key is to look at the products sharing an item_group_id and identify which attributes genuinely distinguish them from each other.

Before you export the feed, review the mapped attributes in Data View and confirm:

When the export runs, the platform combines the individual fields into the structure Google Merchant Center expects for the supplemental feed. For example, a product might produce the following output: