AIVX Labs
AI news for agencies · Getting found in AI answers

Rewriting a product feed for AI Mode with conversational attributes

By Alan Magliocca · Published November 2026

Quick answer: Google's Merchant Center now accepts optional conversational attributes, such as product Q&A, related products and variant options, which it says help AI Mode understand products, and on September 16, 2026 it reported lululemon's attributes were used in half of relevant AI Mode recommendations. Add them through a supplemental data source for each retail client's top products, starting with Q&A pairs written from real customer questions.

Key takeaways

  • There are six conversational attributes: question_and_answer, document_link, related_product, item_group_title, variant_option and popularity_rank.
  • Each product can carry up to 30 Q&A pairs, each question and answer up to 1,000 characters, with 10,000 characters in total.
  • Google recommends adding them through a supplemental data source, and says they won't affect existing products' approval status.
  • Q&A answers must hold product facts only, with no prices, dates or keyword stuffing, and shouldn't repeat the title or description.
  • The 50% figure comes from lululemon's own attributes as reported by Google, not a general result.

What happened

On September 16, 2026, Google's holiday agentic commerce post, by Ashish Gupta, VP and GM of Merchant Shopping, named conversational attributes as one of the feed practices for the season. Google said conversational attributes add richer context to product listings, and that attributes submitted by lululemon were "incorporated 50% of the time in relevant product recommendations in AI Mode." That is one brand's result, reported by Google.

The attributes themselves were not new in September. Search Engine Roundtable reported on May 26, 2026 that Google's help documents for them were live, after Google first previewed them in January. Google's conversational attributes help page says they are optional, designed to complement the main product data specification, and can help products appear on AI-driven surfaces such as AI Mode in Search, as well as in traditional search.

There are six:

  • question_and_answer: frequently asked questions about the product.
  • document_link: links to related PDF documents, such as manuals.
  • related_product: other products in the client's inventory that relate to this one.
  • item_group_title: a title for a product sold in several variants, used with item_group_id.
  • variant_option: the properties that tell variants apart, as name and value pairs.
  • popularity_rank: the product's popularity as a percentage, where higher means it performs better against other products.

What it means for agency owners

AI Mode answers questions like "which tent fits two adults with wide sleeping pads?" A normal feed carries a title, a description and a few specifications. It often can't answer that question directly. Conversational attributes are where you put those answers.

This changes what feed work is. A feed used to be about clean titles, correct prices and approved products. Now part of it is copywriting: writing short, factual answers to the questions shoppers actually ask. That is work agencies already know how to sell.

Two cautions. Google says adding these attributes won't change existing products' approval status, so there is little risk in adding them. But the help pages are strict about content, and wrong answers in a feed are wrong answers in front of shoppers. Every answer needs the client's sign-off.

What are the rules for product Q&A?

Google's question_and_answer page sets the limits:

  • Each pair needs both a question and an answer, each up to 1,000 characters.
  • A product can carry up to 30 pairs, with 10,000 characters in total.
  • Answers hold product information only. Prices and dates belong in other attributes.
  • No keywords or search terms stuffed in.
  • Don't repeat what is already in the title, description, product_detail or product_highlight.
  • If a document_link already points to a manual with the same answers, skip Q&A, because Google says it extracts FAQs from those documents.

Google suggests covering specifications, ingredients, package contents and the occasions or activities the product suits. It also recommends tab-separated files over CSV for text feeds, because answers often contain commas.

For related_product, each entry has three parts: a relationship type, an identifier type (gtin or id) and the identifier. The allowed relationship types are part_of_set, required_part, often_bought_with, substitute, different_brand and accessory. Identifiers may only use letters, numbers, underscores and dashes, and each related product needs its own entry, up to 30.

A worked example: an outdoor store's best-selling tent

Say a retail client sells a two-person backpacking tent, its top product in the tents category. Here is how the rewrite might go, using the client's real specifications in place of the placeholders.

  1. Collect questions. Pull them from the client's product reviews, customer emails and the questions staff hear in the store. Typical ones: does it fit two wide sleeping pads, can one person pitch it, how does it handle rain, what is in the bag.
  2. Write Q&A pairs. Answer each in one or two sentences with the real figure: the floor width in inches, the pitch time the client has tested, the rainfly rating, the items in the stuff sack. Leave out the price and any sale dates.
  3. Add related products. Link the matching footprint as an accessory and the three-person version as a substitute, using the client's product IDs. The text value follows Google's documented pattern of relationship, identifier type and identifier, for example accessory:id:FOOTPRINT-2P.
  4. Set variants. If the tent comes in two colors, give the group one item_group_title and use variant_option to name the color for each variant.
  5. Attach the manual. If the brand publishes a setup PDF, add it as a document_link.
  6. Submit as a supplemental data source, which Google recommends, so the main feed stays untouched.

What to do this week

  1. Pick each retail client's top 20 products by revenue. Don't start with the whole catalog.
  2. Check the popular attributes in Merchant Center's AI performance report to see which details shoppers look for. The 15-minute audit for retail clients shows how.
  3. Gather real questions from reviews, emails and support tickets for those products.
  4. Draft Q&A pairs and related products, get the client to approve every answer, and load them as a supplemental data source using Google's TSV template.
  5. Recheck the AI performance report a few weeks after the change, keeping the same category and time period.

How AIVX Labs helps

AIVX Labs doesn't connect to Merchant Center or edit feeds. It helps with the writing. The Review Analyzer turns a client's reviews into a fix list, and reading those same reviews is a good way to find the real questions shoppers ask. The AI Content Studio can draft answers from specifications you supply, which you then check against the product and the client's records before anything goes into the feed.

Where does this fit with Google's other shopping changes?

Feed quality also matters for checkout inside AI surfaces. If a client is preparing for Google's multi-retailer cart, read what e-commerce clients need for Universal Cart. The same question-first thinking applies to the client's own product pages, covered in making a client's page the one that's linked. For a broader look at how conversational search reads content, BrightStage has a guide to conversational search optimization.

Sources

Written by Alan Magliocca, founder of AIVX Labs and BrightStage AI (Podavinci LLC). Written with AI assistance and checked against the linked sources.

Read our featured article on LinkedIn

Get AIVX Labs verified for $1: RICO, your AI operator, with 7 days of full access.

Get AIVX Labs Verified for $1

Follow along:LinkedInYouTubeTikTokXFacebookInstagramSkool community

Frequently asked questions

Do conversational attributes replace the product description?

No. Google says they complement the main product data specification. Keep the description, and don't repeat what it already says in the Q&A.

Can I put sale prices or holiday dates in the Q&A?

No. Google's question_and_answer rules say to leave out time-related details such as prices and dates, which belong in other attributes.

Will adding these attributes get products disapproved?

Google's help page says including them won't affect the approval status of existing products. Wrong or misleading answers are still a risk to the client, so get every answer approved.

Will every client see lululemon's 50% result?

There's no reason to expect that. The figure describes how often lululemon's own attributes were used in relevant AI Mode recommendations, as reported by Google.