Conversational Attributes in Merchant Center: Optimizing Your Feed for AI Mode (2026)

Adil Ziani

Ecommerce Google Ads Consultant, ex-Google, Google Partner

What Conversational Attributes are

In 2026 Google added a new product data capability to Merchant Center: Conversational Attributes. It exists because shopping queries stopped looking like keywords. Instead of "trail running shoes waterproof", people now ask AI Mode things like "I'm hiking in Scotland in October and my feet always get cold, what should I wear?"

Your classic feed attributes were never designed to answer that. Conversational Attributes are. You add them, along with updated descriptions, directly inside Merchant Center. Google's AI systems then use that structured data to match your products with conversational shopping queries across AI Mode, Gemini and other AI surfaces. If you manage feeds programmatically, the same data is available through the Merchant API under ProductAttributes in the Products sub-API, so you can automate it at scale.

Why keyword-era feeds miss conversational demand

On client accounts I see the same pattern everywhere: feed titles and descriptions optimized for the queries of 2022. Brand, product type, key spec, size, color. That still matters, and I am not telling you to remove it. But it describes what the product is, not what it is for.

Conversational shoppers describe contexts. A gift for a father-in-law who has everything. A stroller that fits in a small Parisian elevator. A moisturizer for someone starting retinol. When your product data says nothing about use cases, occasions or problems solved, the AI has to guess, and it usually guesses a competitor who spelled it out.

How to write conversational attributes that actually match

The goal is not marketing copy. It is honest, structured answers to the questions shoppers actually ask. For each product, I work through four angles:

• Use cases: what the product is used for in real life. "Daily commuting in rain", "meal prep for a family of four", "home office video calls".
• Occasions: when it gets bought or used. Weddings, back to school, first apartment, marathon training, winter travel.
• Who it's for: beginners or experts, sensitive skin, tall riders, small kitchens, new parents. Be specific, not flattering.
• Problems solved: the pain that triggers the search. Overheating laptops, back pain at a desk, dog hair on sofas, frizzy hair in humidity.

Two rules I apply on every account. First, stay truthful: AI systems cross-check your claims against reviews and your product page, and inconsistency costs you trust and visibility. Second, write in natural sentences and phrases, not keyword stuffing. The systems reading this data are language models. They reward clarity.

The workflow: revenue first, not alphabetical

Nobody rewrites a 20,000 SKU catalog by hand, and you should not try. Here is the prioritization I use:

• Pull your last 90 days of product revenue and sort descending.
• Start with the products driving roughly the top 80% of revenue. On most stores that is a surprisingly short list.
• Add conversational attributes and refresh descriptions for those products first, in Merchant Center or through the Merchant API.
• Mine your sources: customer reviews, support tickets and search query reports tell you exactly which contexts and problems people mention.
• For the long tail, use templated attributes by category, then improve them as products earn revenue.

This turns an impossible project into two or three focused working sessions, and it puts the effort where the money already is.

The bonus: this helps beyond Google

Here is the part I like most. The work you do for AI Mode is not locked inside Google. ChatGPT-style assistants build their product understanding from the same raw material: your structured product data and your product pages. Products described with clear use cases, audiences and problems solved get recommended more accurately everywhere, not just on Google surfaces.

So I treat conversational attributes as one investment with multiple payouts: better matching in AI Mode and Gemini, better Shopping ads relevance, and better representation when a shopper asks any AI assistant what to buy.

Where to start this week

Open Merchant Center, pick your ten best sellers, and read their descriptions as if you were an AI trying to recommend them to someone describing a situation. If you cannot tell who the product is for and what problem it solves, neither can the model. Fix those ten first. Measure impressions from AI surfaces over the following weeks, then extend the workflow down your catalog. Feed work is unglamorous, but right now it is the highest leverage hour you can spend on Google visibility.