Google has just published a new set of optional product attributes for Merchant Center product feeds, called conversational attributes. The purpose is to give AI systems and conversational agents, including AI Mode in Search, a structured way to understand the nuances of your products that the traditional product data spec was never designed to express.
This fits Google's broader move toward conversational, intent-driven shopping. Your feed is becoming the knowledge base that AI uses to answer questions, suggest alternatives, and present variants. Conversational attributes are how you feed that knowledge in a clean, machine-readable form.
We support all six new attributes on day one, and we can help with each of them: through AI generation, dedicated modifiers, native parent-product handling, or our reporting engine.
What conversational attributes actually are
The attributes are optional. Submitting them does not affect the approval status of your existing products, and you don't need to duplicate data already provided in description, product_highlight, or product_detail.
There are six attributes:
1) Question and answer (question_and_answer)
A structured FAQ block attached to the product, expressed as question/answer pairs. This is exactly the type of information shoppers ask for in conversational search: "Does this come with a charger?", "Is it dishwasher safe?", "Does it support Bluetooth 6.0?".
If you already publish FAQs on your product pages, that content needs to be promoted into the feed. AI surfaces can then reference or quote your answers directly when answering shopper questions.
How Feedoptimise helps: we can use AI to read your product content, including descriptions, specifications, product pages, and any reference material you import, and build relevant question/answer pairs for each product. The output is composed in the format Google expects, so it can be submitted directly without any further processing on your side.
2) Related product (related_product)
A way to declare structured relationships between products in your catalogue. The attribute is a group of three sub-attributes: relationship type, identifier type, and identifier. Relationship types include accessory, required_part, often_bought_with, and more.
For example, you can tell Google that a specific case is an accessory for a phone listed under a given product ID or GTIN. This goes well beyond what item_group_id or dynamic remarketing can express.
How Feedoptimise helps: we have a dedicated modifier built specifically for related_product. It composes the value in the exact format Google requires, including all three sub-attributes, and lets you build relationships from rules over your catalogue: accessories grouped by category, required parts inferred from product type, often_bought_with derived from performance data or merchandising logic. You can stack multiple rules and produce a comma-separated list of relationships per product.
3) Variant option (variant_option)
Used in combination with item_group_id and item_group_title to describe what makes each variant distinct. Each variant carries a list of name/value pairs such as color:moonstone,memory:512GB,size:8, so AI surfaces can present the variant matrix coherently. This is particularly important for apparel and electronics, where the same product has many SKUs and shoppers ask things like "show me this in black with 256GB storage".
How Feedoptimise helps: we have a dedicated modifier for variant_option that handles the specific name/value structure required by Google. You define which source fields map to which variant axes, including metafields, custom attributes, separate columns, or any combination of these, and the modifier composes the correctly formatted value for every product.
4) Item group title (item_group_title)
The parent title for a family of variants. Where title is the full SKU-level name (for example, "Google Pixel 9 Pro 512GB Moonstone"), item_group_title is the family name ("Google Pixel 9"). It lets AI present a clean product family in conversation, then drill into a specific variant.
How Feedoptimise helps: this one is straightforward. Our native apps and plugins (Shopify, Magento, WooCommerce, Salesforce Commerce Cloud, and others) already expose the concept of a parent product title across the catalogue, so item_group_title maps directly to that value. No additional configuration or content work is required.
5) Document link (document_link)
A URL, or a comma-separated list of URLs, pointing to PDF documents related to the product. Typical use cases are user manuals, assembly instructions, technical datasheets, ingredient lists, and care guides. AI agents can use these documents to answer detailed questions that wouldn't fit inside a description.
How Feedoptimise helps: the mapping itself is simple, the value usually lives in a metafield, a custom attribute, or a dedicated field in your platform. The only requirement on your side is that the documents exist and are publicly accessible. Once that is in place, the field maps directly into the feed.
6) Popularity rank (popularity_rank)
A numeric value representing how a given product performs against the rest of your catalogue, expressed as a percentage. A value of 95.5 means the product sits in the top 4.5% of your inventory by performance. This signal helps AI decide which items to surface when a shopper asks for "your most popular running shoes" or "best-selling winter coats".
How Feedoptimise helps: we can generate popularity_rank directly using our reports functionality. When you connect your feed to a data source such as Google Analytics, Google Ads, or your own store reports, our reporting engine can rank products against each other and produce the percentage value Google expects. The rank then refreshes automatically as performance data updates.
A practical note for retailers and feed managers
Start with the data you already have. If your product pages list FAQs, link to PDF manuals, or expose variant attributes through structured data, most of the work is extraction and mapping rather than content creation. Our native plugins and apps connect to your store through API-like endpoints with full access to your product database, including metafields and custom attributes, so the data is already on hand. For stores without a native integration, our website crawler can extract the same information from product pages.
Where Google Shopping is heading
Google has been steadily reorganising Shopping around AI Mode, Business Agent, and conversational discovery. The signal is consistent: structured product knowledge is what AI needs in order to recommend, compare, and present your products inside conversational surfaces. Conversational attributes are the first standardised vocabulary Google has shipped for that purpose.
If you want to start submitting them, you can do it today inside your Feedoptimise account. If you'd like help with AI-generated Q&A, structured relationships, variant options, or popularity ranking from your performance data, our team can set it up with you.