Google Merchant Center Conversational Attributes Guide

If you sell on Shopify or Magento, start rebuilding your product data now, at the catalog level, before Google finishes rolling out Google Merchant Center conversational attributes across AI Mode shopping. Don't wait for the final spec, and don't patch the problem inside a feed tool. AI Mode answers questions such as "which insulated bottle fits a car cup holder and survives a dishwasher," and it can only answer from facts your feed actually contains. Stores that already hold complete, structured, specific product facts will be ready on day one. Stores that hold a title, a price, and a marketing paragraph will be skipped. The fix lives in your catalog, and it takes weeks of data work.

What conversational attributes ask of your feed

Classic Shopping ads matched keywords to titles and categories. A shopper typed "stainless water bottle 32 oz," and Google matched words in your title and product type. Conversational shopping works differently. The shopper describes a situation in full sentences, adds constraints, and follows up. The system has to break that request into attributes and filter products against them: capacity, material, dishwasher safety, lid type, fit, use case, compatibility, return terms.

A product can only pass that filter if the attribute exists as a discrete, machine-readable value. "Perfect for your daily adventures" answers nothing. "Capacity: 32 oz. Dishwasher safe: lid only. Fits standard car cup holders: no" answers three follow-up questions.

That's the whole game. Conversational attributes reward feeds that state product facts the way a knowledgeable salesperson would answer a customer, one fact per field, with no adjectives standing in for data.

Google's documentation for these attributes is still moving, and we won't restate field names here that we can't verify against the current spec. You don't need the final list to start. Every useful conversational attribute rests on facts you can already supply through fields that exist in Merchant Center today:

• title and description, written around attributes instead of slogans
• product_type and google_product_category, mapped precisely
• brand, gtin, and mpn, so Google can match your product to its product graph
• color, size, material, pattern, age_group, and gender where they apply
• product_detail, which accepts structured section, attribute, and value groupings for technical specs
• product_highlight, for short, factual selling points
• shipping, return policy, and availability data that stays accurate

Stores that fill these completely and accurately will map cleanly onto whatever conversational fields Google adds next, because those fields describe the same underlying product facts.

Fix the catalog, not the feed tool

When a feed underperforms, the usual reflex is to add a feed management layer and rewrite titles with rules. Feed rules have a place. They fall apart when the data they need doesn't exist upstream.

A rule can append "32 oz" to a title only if capacity lives in a field somewhere. It can't invent dishwasher safety, compatibility, or care instructions. If those facts sit inside an unstructured description, or nowhere, the rule has nothing to work with.

There's a second problem. AI Mode doesn't read your feed in isolation. It cross-checks your feed against your product page, your structured data, and everything else it knows about the product. A feed tool that writes "dishwasher safe" into the feed while your product page says nothing about it creates a mismatch. A feed tool that rewrites "hand wash only" into something friendlier creates a disapproval risk and a returns problem.

So the work belongs in the catalog:

1. Define the attributes that matter for each product category.
2. Create real attribute fields in Magento or Shopify for each one.
3. Populate them product by product.
4. Feed Merchant Center, product page templates, and schema markup from those same fields.

One source of truth, three outputs. When the data changes, all three change together.

What our AI Overview data says about where to start

Shopping surfaces don't run on organic rank alone, but the organic foundation still decides a lot of what AI systems cite. In a July to September 2026 analysis of client sites we manage, we looked at 2,350 tracked keyword snapshots where a Google AI Overview appeared, covering 314 keywords. The AI Overview cited the site 73% of the time when that site held organic position 1. Citation fell to 49% at positions 2 and 3, 42% at positions 4 through 10, 9% at positions 11 through 20, and 3% when the site ranked beyond position 20 or didn't rank at all. The pattern held on nearly 80% of the sites. The rest didn't have enough data to judge. We broke this down in AI Overview Citations Follow Rank: What Our Data Shows.

That's a correlation, not proof of cause. Two details from the same analysis matter directly for product feeds, though.

First, about a quarter of citations went to a different page on the same site than the page that ranked. A category page ranks, and the AI cites a product page, or the reverse. Your product pages need to stand on their own as answers, with the same structured facts your feed carries.

Second, the citations that showed up without a top-20 rank were mostly comparison queries. Comparison is exactly what conversational shopping does: "this bottle versus that one for hiking." A store can earn visibility on those queries without ranking first, provided its product facts are complete enough to compare. Clean attributes are how a mid-ranked store competes.

So the priority order looks like this. Protect and improve organic rank on your money categories. Make every product page a self-contained answer. Make every attribute comparable.

The attribute audit we run before touching the feed

Before writing a single feed rule, we audit the catalog against the questions shoppers ask in each category. Here's the sequence.

1. Pull the question set per category. Use Search Console queries, site search logs, customer service tickets, and product reviews. Reviews are the most honest source. Every "I wish I'd known it didn't fit my cup holder" review points to an attribute you should have published.

2. Turn questions into attributes. "Does it fit a cup holder" becomes a base diameter field. "Will it keep ice overnight" becomes an insulation type field and a temperature retention claim, which you should publish only if you can back it up. "Can I put it in the dishwasher" becomes a care field with specific values for body and lid.

3. Score fill rate per attribute per category. A field that's populated on 30% of products might as well not exist for filtering purposes. Track fill rate by category, because a 90% fill rate overall can hide a category sitting near zero.

4. Check value consistency. "Stainless," "stainless steel," "18/8 SS," and "steel" are four values for one fact. Normalize them to one controlled vocabulary. AI filtering treats inconsistent values as different things.

5. Check feed-to-page parity. Every attribute in the feed should appear on the product page, ideally in a visible spec table that also renders as Product structured data.

6. Check identifiers. Missing or wrong GTINs keep Google from connecting your listing to the product it already understands. For brands that carry other manufacturers' products, this is often the single biggest gap.

7. Check offer data. Price, availability, shipping cost, delivery estimates, and return terms. Conversational shoppers ask "can I get it by Friday" and "can I return it." Offer accuracy also feeds into Google's agentic checkout work, which we covered in Google Universal Cart for Retailers: Feed, Checkout, SEO.

A product example, for illustration only, shows the difference. Here's a typical feed entry for an insulated bottle:

Title: Adventure Bottle - Blue Description: Stay hydrated in style with our premium insulated bottle, built for every adventure.

Here's the same product rebuilt for conversational queries:

Title: Adventure Insulated Stainless Steel Water Bottle, 32 oz, Wide Mouth, Blue Material: 18/8 stainless steel Capacity: 32 oz Lid type: wide mouth, straw lid included Care: body hand wash; lid top-rack dishwasher safe Fits standard car cup holder: no product_highlight: Double-wall vacuum insulation; BPA-free lid; wide mouth fits standard ice cubes

The second version answers six questions the first one ignores. None of it is marketing copy, and all of it can be filtered.

Magento: where product feeds break

We've inherited a lot of Magento stores over the years, and the feed problems on them usually come from the same few places.

Extension sprawl. We regularly find 40 to 70 extensions on stores that need maybe 15. We've opened stores with 60+ installed. When three of those extensions touch product attributes, a feed export, and structured data, nobody can say which one controls what. Feed values get overwritten on export, schema contradicts the feed, and fixes made in one place get undone by another. Before you optimize a Magento feed, inventory every extension that reads or writes product data and cut it down to one owner per job.

Attribute sets built for display, not data. Magento's attribute system is strong. On a lot of stores, though, someone created attributes to drive a product page layout, not to describe the product. You end up with a "features" text area holding ten facts in one blob. Split those into discrete attributes with controlled values, assign them to the right attribute sets, and map them to feed fields.

Configurable products and variant data. Google needs variant-level data: each color and size as its own item with its own identifiers, linked by item_group_id. If your configurable product exports only parent-level data, or exports children without their distinguishing attributes, conversational filters can't separate the 24 oz from the 32 oz.

Orphaned and broken product URLs. One technical SEO audit we ran found crawl issues on 12,000 orphaned product URLs. Orphaned pages are a problem for any feed, because a link that Google can't crawl or that sits outside your site structure weakens the page-level confirmation AI systems look for. Make sure every product in the feed links to an indexable, internally linked page.

Magento 1. We still find stores running Magento 1, completely unpatched. Magento 1 reached end of life in June 2020 and gets no security patches. Treat any store still on it as already compromised. Feed optimization on a Magento 1 store is wasted effort. The move to Magento 2 comes first, and catalog cleanup should be part of that migration rather than a follow-up project.

Magento rewards stores that treat the catalog as structured data from the start. That's how we approached the Magento build for WaterBottles.com, where we handle e-commerce strategy, the build, and SEO under one team.

Shopify: where product feeds break

Shopify stores fail differently. The platform keeps the core tidy, and the gaps show up at the edges.

Facts trapped in descriptions. Shopify's default product model is title, description, variants, and a few options. Everything else ends up in the description body. Move product facts into metafields with defined types, and use Shopify's standard product taxonomy and category metafields where they fit your products.

Option limits forcing workarounds. When a product needs more variant dimensions than options allow, stores split products or bundle attributes into option names like "32 oz / Straw Lid / Blue." That breaks clean attribute mapping. Fix the structure so each fact lives in its own field.

App overlap. Shopify feed problems often trace back to two or three apps writing to the same data, much like Magento extension sprawl. Pick one channel integration path to Merchant Center, and make sure your metafields map to it explicitly instead of relying on defaults.

Theme spec tables. A lot of Shopify themes have no structured spec section on the product page at all. Add one that pulls from metafields, so the page confirms what the feed states.

Paid and organic run on the same feed

Your Merchant Center feed powers free listings, Shopping ads, Performance Max, and the shopping surfaces inside AI Mode. That means feed work is paid media work and SEO work at the same time. A store that lets a paid team edit feed titles for click-through while an SEO team edits product pages for rank ends up with two versions of the truth and nobody accountable for either.

This is why we keep strategy, SEO, and development on one team. The person deciding which attributes matter for AI Mode visibility also decides how they're built in Magento and how they appear in campaigns. If you want the paid side handled alongside the catalog rebuild, our Google Ads management services work from the same feed and the same attribute decisions. For how ad formats are changing inside AI Mode, read Google Ads Direct Offer AI Mode: How to Prepare Now.

Why LA brands should move first

Los Angeles e-commerce runs heavy in categories where conversational queries do the most damage to thin feeds: apparel, beauty, supplements, home goods, outdoor gear, and accessories. These categories are full of fit, material, ingredient, and compatibility questions. They're also crowded, with local and national brands competing on the same products.

In a crowded category, the store with the more complete attribute set wins comparison queries even without the top organic position. That's the opening our data points to. The brands that do the catalog work first set the bar that everyone else gets compared against.

Where we could be wrong

Google could ship conversational attributes that depend mostly on its own extraction from product pages and reviews, with feed fields carrying less weight than we expect. If that happens, the catalog work still pays off, because the same structured facts on your product pages become what Google extracts. The only scenario where this work goes to waste is one where Google stops caring about product facts entirely, and nothing in how AI Mode answers shopping questions points that way.

The other risk is timing. If Google's rollout moves slower than we expect, you'll have finished early. Finishing a catalog cleanup early isn't a cost. The same data improves your current Shopping campaigns, free listings, and on-site filtering today.

A 30-day plan to get ahead

Week 1: Audit. Pull Merchant Center diagnostics, export the full catalog, and score attribute fill rate by category. Build the question set per category from Search Console, site search, reviews, and support tickets. Inventory every Magento extension or Shopify app that touches product data.

Week 2: Define. Pick your top revenue categories. For each one, define the attribute list, the controlled vocabulary for each attribute, and the mapping to Merchant Center fields. Decide which system owns each field, and remove the duplicates.

Week 3: Build. Create the attribute fields in Magento or the metafields in Shopify. Add a spec table to product page templates that renders the same fields as Product structured data. Fix variant exports and identifiers.

Week 4: Populate and ship. Fill attributes for your top categories, starting with the highest-revenue SKUs. Rewrite titles around attributes. Push the feed, check Merchant Center for disapprovals and warnings, and spot-check feed-to-page parity on a sample of products.

Then repeat for the next set of categories. Measure progress by fill rate per category, disapproval rate, and impression share across Shopping and free listings. Make sure AI-referred traffic gets attributed correctly in GA4. Our guide to ChatGPT traffic showing as direct in GA4 covers how to sort that out.

Get your feed audited

The stores that show up in AI Mode shopping answers will be the ones whose catalogs answer the questions shoppers ask. That work takes Magento or Shopify development, SEO, and feed strategy done by one team, and it takes weeks rather than days. Start before your competitors do.

Get an audit. We'll score your catalog's attribute coverage by category, find where your Magento extensions or Shopify apps are fighting over product data, and hand you a prioritized rebuild plan. If you're ready to move, request a proposal and we'll scope the catalog work, the feed, and the campaigns together.