Google Merchant Center Conversational Attributes: What They Are and How to Actually Use Them for AI Mode

If you sell anything through Google Shopping, there's a decent chance you missed this one. Google didn't put out a press release or a keynote slide for it. It just quietly updated the Merchant Center product data spec with a new group of fields, tucked under the label "conversational attributes," and left a fairly dry support page to explain them.

But if you actually work in ecommerce or SEO right now, this update matters more than most of the flashier AI announcements from the last year. It's Google telling merchants, in plain terms, how it wants product data formatted so that AI Mode, Gemini, and other conversational shopping experiences can actually understand what you sell — not just index it.

Here's what changed, why it exists, and what to do about it before your competitors figure it out first.

What are conversational attributes, exactly?

Conversational attributes are six new, fully optional fields you can add to your Merchant Center product data:

• question_and_answer — FAQs about the product, written as literal question/answer pairs

• document_link — links to PDFs like manuals, spec sheets, or assembly instructions

• related_product — structured relationships between products (accessories, required parts, frequently bought together)

• item_group_title — a shared title for a group of product variants

• variant_option — the specific attributes that distinguish one variant from another (size, color, memory, width)

• popularity_rank — how well a product is performing relative to the rest of your catalog, expressed as a percentage

None of these are new categories of information, exactly. Most stores already have FAQ sections, spec sheets, and variant selectors on their product pages. What's new is that Google now wants this information submitted in a structured, machine-readable format specifically so AI systems can pull it directly instead of trying to scrape and interpret your website.

Google's own framing is worth repeating: these attributes exist to "help AI systems and conversational agents better understand your products' specific nuances." That's a very deliberate phrase. Traditional Shopping feeds were built to answer questions like "what is this" and "how much does it cost." Conversational attributes are built to answer "does this work for me," which is a completely different kind of query — and increasingly the kind of query people are typing (or speaking) into AI Mode.

Why this is happening now

Think about how someone searches inside AI Mode versus how they used to search on Google. A classic Shopping query looks like wireless earbuds under $100. An AI Mode query looks more like wireless earbuds that don't have a headphone jack issue and work well for running.

That second query can't really be answered by a title, a price, and a GTIN. It needs context: does it have Bluetooth, is it sweat-resistant, what's actually in the box, what do people usually buy alongside it. Historically, a lot of that context lived buried in a paragraph on the product description page, or in a comparison table your dev team built and Google's crawler may or may not have parsed correctly.

Conversational attributes remove the guesswork. Instead of hoping an AI model correctly extracts "no headphone jack" from three paragraphs of marketing copy, you just tell it directly:

That's the whole point. It's less about ranking and more about retrieval accuracy — giving the model a clean, unambiguous source to cite instead of forcing it to infer.

Should you actually fill these out?

Google is explicit that these attributes are optional and won't affect your existing product approvals. So the honest answer is: you don't have to touch any of this, and nothing breaks if you ignore it.

But "optional and low-risk" is usually exactly the kind of update worth doing early, before it becomes table stakes. A few reasons this is worth prioritizing over the next quarter rather than filing it under "someday":

AI Mode is already routing purchase-intent traffic. If your competitor's product includes a clean answer to "does this fit a 6'2" frame" and yours doesn't, guess whose product gets surfaced and cited when someone asks that exact question.

It's genuinely low effort if you're organized. Most of this data already exists somewhere in your product catalog, your customer support macros, or your existing FAQ content. This is largely a formatting and submission exercise, not a content creation project from scratch.

Popularity rank is a rare gift. Google almost never lets merchants tell it directly which of their own products perform best. This attribute is essentially Google asking you to hand it a hint about what to promote. Ignoring a free ranking signal seems like a mistake.

Related products fix a long-standing blind spot. The related_product attribute lets you explicitly tell Google which item is a required accessory versus which one is frequently bought together, using defined relationship types. That's the kind of cross-sell logic that used to live entirely in your own site's recommendation engine, invisible to Google.

A practical way to roll this out

You don't need to boil the ocean here. A reasonable rollout looks something like this:

1. Start with your best sellers. Pick your top 20–50 SKUs by revenue and build out full conversational attribute sets for those first, rather than trying to backfill your entire catalog at once.

2. Mine your existing support tickets and FAQ pages for question_and_answer content. If customers keep asking the same three questions about a product, that's your starting list.

3. Use a supplemental data source, which Google recommends over jamming everything into your primary feed. It keeps your core feed clean and lets you iterate on conversational data separately.

4. Don't duplicate what's already in description, product_highlight, or product_detail. Google explicitly says not to repeat information you've already submitted elsewhere — the goal is additive nuance, not redundancy.

5. Treat variant_option and item_group_title as a pair. If you sell products in multiple sizes, colors, or configurations, this combination is what lets an AI system understand "the phone" and "the phone in moonstone, 512GB" as the same underlying product with different real options — which matters a lot when someone asks a conversational follow-up like "does that come in a smaller size."

The bigger pattern here

Zoom out and this update is part of something broader that anyone doing SEO right now has probably already noticed: search is splitting into two disciplines. There's still classical SEO — titles, structured data, backlinks, page speed. And then there's a newer, separate discipline that's really about being legible to language models: writing content that answers questions directly, structuring data so it can be retrieved cleanly, and giving AI systems unambiguous facts to cite instead of vague marketing prose.

Conversational attributes are Google formalizing that second discipline for ecommerce specifically. It's the same underlying shift you're seeing with schema markup for AI citations, llms.txt experiments, and the general move toward structured, retrieval-friendly content across the web. Merchant Center is just where it's showing up for product data.

The merchants who treat this as a checkbox exercise will fill in a few fields and move on. The ones who actually think about what their customers ask before buying — and answer those questions directly, in the format Google is now asking for — are going to be the ones showing up when someone asks AI Mode a real question instead of typing a keyword.

That's a small shift in behavior for a potentially large shift in visibility. Worth the afternoon it takes to set up.

Quick answers to what people are asking about this update

What are conversational attributes in Google Merchant Center? They're six optional product data fields — question_and_answer, document_link, related_product, item_group_title, variant_option, and popularity_rank — designed to help AI Mode, Gemini, and other AI shopping surfaces understand product details that a standard title-and-price feed can't capture.

Are conversational attributes required? No. Google has confirmed they're completely optional and won't affect the approval status of existing products. They're additive, not a compliance requirement.

Where do I add conversational attributes? Either directly in your primary Merchant Center data source, through a supplemental data source (Google's recommended approach), or via the Merchant API.

Do conversational attributes replace product descriptions or product highlights? No, and you shouldn't duplicate content between them. If a detail is already covered in description, product_highlight, or product_detail, Google says you don't need to repeat it in the conversational fields.

What is the popularity_rank attribute used for? It's a self-reported percentage showing how a product performs relative to the rest of your catalog, letting you signal your own best sellers directly rather than relying on Google to infer it from sales and click data alone.

Will this affect my regular Google Shopping rankings, not just AI Mode? Google positions these attributes as also enhancing traditional search experiences, not just AI-driven ones, so there's some crossover benefit even outside of AI Mode specifically.