Google has added a new section to the Merchant Center product data specification: Conversational Attributes. Six new optional attributes were added to the Help Centre immediately following Google Marketing Live 2026, which give retailers a structured way to provide additional product information specifically designed for AI-driven surfaces—AI Mode in Search, AI Overviews, conversational agents, and more.
The attributes are optional. They do not affect the approval status of existing products. And they are designed to sit alongside existing feed data rather than replace it. Google explicitly states that if you already submit specific details in [description], [product_highlight], or [product_detail], you do not need to duplicate that data in the conversational attributes.
What they represent is a new input layer between retailer product data and AI Search. As Google builds conversational shopping experiences like AI Mode, Universal Cart, and Direct Offers, the quality and structure of the data those systems draw on becomes the upstream determinant of whether a product surfaces, how it surfaces, and how accurately it is represented to a shopper. Conversational Attributes are how Google is asking retailers to enrich that data.
AI systems can only recommend what they understand. Conversational Attributes are how you tell them.
The Six Attributes
Attribute 01: Question and Answer
[question_and_answer]
A structured Q&A field where retailers add anticipated customer questions and their answers directly against the product record. This is the most directly useful attribute for conversational AI surfaces. AI Mode handles a high volume of exploratory, question-based queries, and a product with pre-answered questions is structurally better positioned to match those queries than one that relies on copy-extracted inference alone.
Use this for the questions your product pages already answer in FAQs or customer service interactions. Does it have a headphone jack? What are the dimensions? Is it compatible with X? What size should I order?
The attribute accepts multiple Q&A pairs in a single field.
Example value:
“Does it have a headphone jack?”:“This version doesn't have a headphone jack.”, “Does it support Bluetooth?”:“It has full Bluetooth 6.0 support.”
Attribute 02: Document Link
[document_link]
A URL field for linking product-related PDF documents: manuals, assembly instructions, spec sheets, care guides, safety documentation. Accepts multiple PDF URLs separated by a comma.
For categories where specification depth matters, such as electronics, appliances, furniture, tools, or anything with a manual, this attribute gives AI systems access to structured product documentation that would otherwise require crawling the product page or inferring from copy. A user asking detailed installation or compatibility questions is better served when the AI has access to the actual documentation.
Example value:
https://example.com/manual.pdf, https://example.com/assembly_instructions.pdf
Attribute 03: Related Product
[related_product]
A group attribute that defines relationships between products, accessories, required parts, and frequently bought together items. Uses three sub-attributes: relationship type, identifier type, and identifier.
This attribute is particularly relevant to Universal Cart and agentic shopping experiences where AI assembles product recommendations. A retailer who explicitly signals which products are accessories or required components gives the AI the relational data it needs to surface complete solutions rather than isolated products. It also directly informs the “often bought with” recommendation logic that drives basket size.
Example value:
required_part:id:AZ7B, accessory:gtin:811571013579, often_bought_with:id:AZ7C
Attribute 04: Item Group Title
[item_group_title]
Assigns a human-readable title to a product variant group. The parent product name that sits above individual variant records. Used in combination with the existing [item_group_id] attribute.
Where [item_group_id] is a machine-readable identifier, [item_group_title] is a label AI systems can use in natural language responses. When a conversational agent discusses a product family — “the Google Pixel 9 range” — having a consistent group title prevents fragmented or inconsistent naming across variant records.
Example value:
Google Pixel 9
Attribute 05: Variant Option
[variant_option]
Specifies all variant-identifying properties of a product when it is available in different variants. Used in combination with [item_group_title] and [item_group_id]. Accepts name and value sub-attributes, allowing multiple variant dimensions in a single field.
Standard feed attributes like [colour] and [size] cover common variant types, but many categories have product-specific variant dimensions that don't map cleanly to those fields, such as shoe width, memory configuration, display size, and blade length. [variant_option] provides a flexible field for declaring those dimensions in a structured way that AI systems can parse and use in conversational responses.
Example value:
display:XL,memory:512GB,color:moonstone
Attribute 06: Popularity Rank
[popularity_rank]
A numeric value indicating the popularity of a product as a percentage of total inventory. The higher the value, the better the product performs relative to other products in the retailer's catalogue. Submitted as a decimal between 0 and 100.
This attribute gives AI recommendation systems a retailer-supplied popularity signal that sits alongside Google’s own performance data. For categories with large catalogues, it helps AI surfaces prioritise products the retailer knows to be strong performers rather than relying solely on inferred signals. It is also the most straightforward of the six to implement: a single numeric value derived from sales or conversion data that most retailers already track internally.
Example value:
95.5
How To Implement The Six Attributes
Google recommends adding conversational attributes via a supplemental data source rather than your primary feed. This keeps the new attributes separate from existing product data, makes them easier to maintain and update independently, and avoids any risk of disrupting your primary feed approval status.
They can also be submitted via the Merchant API.
The practical implementation order for most e-commerce brands depends on catalogue size and category. For most retailers, the highest-leverage attributes to implement first are:
Question and answer first — it directly addresses the query format AI Mode handles, and most retailers already have this content in FAQ sections or customer service scripts. The work is structuring and submitting existing content, not creating new content.
Related product second — if you have accessory and complementary product relationships that currently only live on product pages, submitting them via [related_product] makes them accessible to AI shopping agents assembling product recommendations.
Popularity rank third — a single numeric value per product, derivable from existing sales data. Low implementation cost, direct signal to AI recommendation systems.
What not to duplicate
Google is explicit: if you already submit product information in [description], [product_highlight], or [product_detail], you do not need to repeat that data in conversational attributes. The six new attributes are additive. They are designed to capture information that does not fit cleanly into existing feed fields, not to replicate what is already there.
The immediate priority
Question and answer is the highest-priority attribute to implement for most e-commerce brands running Google Shopping. AI Mode queries are conversational and question-based by nature. The format of AI search is a question seeking a specific answer. Products that have pre-structured Q&A data are providing exactly the signal those systems need. The content almost certainly exists in your business already. The work is extracting it from your FAQ or customer service resources and structuring it in the feed.
Why The Timing Matters
Conversational Attributes were announced at Google Marketing Live 2026 on 20 May and added to the Merchant Center Help Centre immediately after. They are not a future capability. They are available now for implementation.
The context for why they matter is what is happening to the product discovery surface. AI Mode has surpassed 1 billion monthly users with queries doubling every quarter. Google’s new ad formats, like Conversational Discovery ads, Highlighted Answers, AI-powered Shopping ads, all draw on product feed data to generate AI-written responses and summaries per query. Universal Cart is live in the United States and expanding to the UK in the coming months. Direct Offers is assembling product bundles via Gemini from retailer-supplied promotions and guardrails.
Every one of those surfaces is making decisions about which products to surface, how to describe them, and how to match them to a user's specific question. The quality of that decision is constrained by the quality of the data. Conversational Attributes are a mechanism for retailers to make that data richer, more specific, and more useful to AI systems before those systems make the surfacing decision.
The Honest Caveat
These attributes are optional and Google does not specify precisely how they weight in AI surfacing decisions. There is no published evidence yet of specific performance uplifts from implementing them. They are new to the spec. What is clear is the direction: Google is building AI shopping surfaces that use product data as their primary input, and it is giving retailers a more detailed vocabulary for describing their products to those systems. Early implementation has no downside risk. The attributes do not affect product approval status, and they can be added via a supplemental data source without touching the primary feed.
Six new optional attributes in the Merchant Center product data spec. All six designed to give AI systems more structured product data to work with across AI Mode, AI Overviews, conversational agents, and Google's expanding agentic commerce surfaces. Implementation via supplemental data source, no effect on existing product approval status.
The full specification for each attribute is available directly in the Google Merchant Center Help Centre at the links in the sidebar.
Need a fresh perspective? Let’s talk.
At 360 OM, we specialise in helping businesses take their marketing efforts to the next level. Our team stays on top of industry trends, uses data-informed decisions to maximise your ROI, and provides full transparency through comprehensive reports.
.png)
