Why Did Google Introduce Conversational Attributes To Shopping Feeds In May 2026?
2.5 billion people use Google AI Overviews every month.
AI Mode has already surpassed 1 billion monthly users.
And just one year after launching, AI Mode queries are more than doubling every quarter.
Google says its new AI features are a leading reason Search queries have reached an all-time high.
For ecommerce brands, there's another number worth paying attention to.
A 2026 Semrush analysis of more than 600,000 keywords found that the presence of AI Overviews on searches with commercial intent grew 71% in six months.
This isn't a side experiment inside Google anymore.
LLMs is becoming part of how people search. And increasingly, how they shop.
From Keywords to Conversations
For the last two decades, ecommerce marketers have become exceptionally good at optimizing for keywords.
A shopper searches:
“black polarized sunglasses”
Google matches that query against product titles, descriptions, categories, attributes, landing pages, bids, and dozens of other signals.
Your Shopping feed is built for this world.
And optimizing it for traditional Shopping should still be priority #1.
But now consider this search:
“I'm going to Portugal this summer and need sunglasses for the beach. I have a round face and want something lightweight and polarized. What should I buy?”
That's not really a keyword search.
It's a conversation.
Google needs to understand more than whether a product is categorized as “sunglasses” and has “black” somewhere in its title.
It needs to understand:
Is this product good for round faces?
Is it lightweight?
Are the lenses polarized?
Does it reduce glare at the beach?
Who is this product best suited for?
How does it compare with other options?
Would it actually be a good recommendation for this shopper?
And Google is rapidly building Search around exactly these kinds of interactions.
AI Overviews now answer questions directly on the search results page. AI Mode allows users to go deeper, ask follow-up questions, compare options, and continue searching without starting over.
Google is also connecting the two experiences. A shopper can begin with a traditional Google search, receive an AI Overview, and continue directly into AI Mode while preserving the context of the original query.
The search box is becoming a conversation.
And that creates a new challenge for anyone responsible for a Shopping feed:
How do you optimize product data for questions instead of just keywords?
Google has started giving merchants an answer.
Enter Conversational Attributes
In 2026, Google introduced a new set of optional Conversational Attributes in Merchant Center.
According to Google, these fields are designed to help AI systems and conversational agents understand the nuances of a product and provide better information across AI-powered experiences.
There are currently six:
- Question & Answer
- Document Link
- Related Product
- Item Group Title
- Variant Option
- Popularity Rank
Google's documentation is worth reading directly:
[Google's Conversational Attributes Guide]
[Google's Question & Answer Specification]
We've also created a [long-form video walkthrough of Conversational Attributes] for anyone who wants to go deeper into the broader update.
But for this article, we're focusing on one field:
Question & Answer (question_and_answer).
Because Google gives you something interesting here.
Up to 30 questions and answers for every SKU.
And we think most brands are going to approach those 30 questions the wrong way.
30 Questions Doesn't Mean 30 FAQs
The easy approach looks like this:
Take your PDP.
Put it into an LLM.
Prompt:
“Generate 30 FAQs about this product.”
Upload them to Merchant Center.
Done.
But what have you actually accomplished?
If Google gives you 30 opportunities to provide additional context about a product, the goal shouldn't be to fill 30 fields.
The better question is:
Which 30 questions could connect this product to the greatest number of relevant shopping conversations?
That's the question we've been working on at Banfana.
And after building these feeds with brands like CRAP Eyewear, we've developed a framework for approaching it:
PBC² → PULSE → The Top Product Rule
PBC² tells you what kinds of questions to ask.
PULSE tells you where to find the best questions.
The Top Product Rule tells you where to concentrate them.
Let's break it down.
Step 1: PBC² — Classify the Questions
We start by breaking potential questions into four territories:
P — Product
These are questions specifically about the SKU.
“What material are these frames made from?”
B — Brand
Next are questions shoppers ask about the brand.
For example:
“What makes CRAP Eyewear different from other sunglasses brands?”
C — Competitor
Then we have competitor and alternative questions.
For example:
“Is there good alternative to Ray-Ban?”
C — Category
Finally, we have broader category questions.
These are often questions beginning with what, why, how, top, or best.
In the CRAP Eyewear feed we've built, examples include:
“What are polarized sunglasses?”
“What shape sunglasses are best for a round face?”
These questions may not mention the brand or SKU at all.
That's the point.
They represent broader conversations where the product could be a relevant answer.
That's PBC²:
Product. Brand. Competitor. Category.
Now comes the harder question.
How do you decide which of those questions deserve one of your 30 slots?
Step 2: PULSE — Don't Brainstorm Questions. Mine Them.
Anyone can brainstorm 30 questions.
We want evidence that shoppers actually care about them.
That's where PULSE comes in.
P — Product Truth
Start with the product itself.
Pull from your PDP, existing feed attributes, specifications, materials, dimensions, sizing, lens details, product documentation, and other first-party information.
For example, if a pair of CRAP Eyewear sunglasses uses polarized CR-39 lenses with 100% UVA/UVB protection, we have the factual foundation needed to answer questions around polarization and protection.
U — User Voice
Next, listen to your customers.
We're building our process to ingest consumer feedback from sources including:
- Website reviews
- Amazon reviews
- Customer feedback
The goal isn't simply to summarize sentiment.
We're looking for repeated needs, use cases, attributes, objections, and unexpected benefits.
Maybe customers repeatedly mention that a frame works well on wider faces.
Maybe they talk about wearing it while driving.
Maybe they love how lightweight it feels.
L — Long-Form Content
Then we look beyond written reviews.
YouTube and TikTok reviews can contain some of the richest product feedback available.
By analyzing transcripts, we can identify how creators and customers naturally talk about:
- Fit
- Style
- Comfort
- Use cases
- Comparisons
- Pros and cons
- Who should buy the product
This matters because that language often looks much more like an AI Mode conversation than traditional keyword research does.
S — Search Demand
Then we look at what people are actually typing into Google.
Two of our most important sources are:
Google Ads Search Terms Reports
and
Google Search Console.
This is where the framework gets particularly interesting.
In the search data behind our CRAP Eyewear work, we found queries around topics like:
“what are polarized sunglasses”
“polarized vs non polarized sunglasses”
“alternative sunglasses”
“beach sunglasses”
Instead of guessing which category questions shoppers might care about, we can use actual search behavior to guide the feed.
If people are already asking Google “what are polarized sunglasses?” and a product is genuinely polarized, that's a strong candidate.
The answer can explain the concept and connect it directly back to the SKU:
A polarized lens reduces glare from reflective surfaces such as roads, water, and sand and this particular product uses polarized CR-39 lenses with UVA/UVB protection.
One question now connects category education with product relevance.
E — Evidence
Finally, we look for convergence.
The strongest questions tend to sit at the intersection of:
Product Truth × Consumer Language × Search Demand
Suppose:
Search data shows people looking for “sunglasses for round faces.”
Customer feedback repeatedly discusses how a particular frame fits.
And the product data tells us that the frame has an angular square silhouette.
Now we have evidence for a question like:
“What shape sunglasses are best for a round face?”
And the answer can explain the broader category principle while showing why this particular SKU fits the consideration set.
That's fundamentally different from asking ChatGPT:
“Give me 30 FAQs about these sunglasses.”
The best conversational attributes aren't invented.
They're discovered.
Step 3: The Top Product Rule
There's one final question:
Should every SKU receive Product, Brand, Competitor, and Category questions?
Our answer is no.
All Products → Product Questions
Your catalog should have strong product-level coverage.
The objective is simple:
Make every product understandable.
Give Google the additional context it needs to accurately understand the nuances of each SKU.
Top Products → PBC²
Your Top Products, however, should receive broader coverage.
Product.
Brand.
Competitor.
Category.
What counts as a Top Product depends on the business.
It could include:
- Best sellers
- Hero products
- Highest-margin products
- Strategic growth products
- Products with strong inventory
- Products the merchandising team wants to push
Why prioritize?
Imagine someone asks:
“What are the best festival sunglasses for women?”
You probably don't need every women's frame in your catalog positioned around that conversation.
You want the products that are actually relevant — and that the business most wants to sell.
Or imagine someone asks:
“What are some good alternatives to Oakley?”
Your classic oval fashion frames probably don't belong there.
Your strongest sport-ready wraparound frames might.
Conversational feed optimization isn't about associating every product with every possible question.
It's about identifying which conversations each product deserves to be part of.
That's the Top Product Rule.
Make every product understandable. Make your top products discoverable.
The Shopping Feed Is Becoming a Product Knowledge Base
Traditional Shopping feed optimization isn't going away.
Titles still matter.
Product types still matter.
GTINs, colors, sizes, materials, descriptions, images, pricing, availability, and the rest of the feed still matter.
That remains the foundation.
But Google is giving us a clear indication of where Search is heading.
In April 2026, Google introduced AI Max for Shopping by making almost this exact point: shoppers aren't only searching for products anymore. They're asking questions like:
“What are the best high-quality clothes for lounging?”
That's a different kind of search.
And it requires a different depth of product information.
Conversational Attributes give merchants an opportunity to start building that depth directly into the Shopping feed.
Our framework is straightforward:
PBC²
What conversations could this product belong in?
Product → Brand → Competitor → Category
PULSE
Which questions are actually worth answering?
Product Truth → User Voice → Long-Form Content → Search Demand → Evidence
The Top Product Rule
Where should you concentrate the broadest coverage?
All Products → Product Questions
Top Products → PBC²
The objective isn't to fill 30 fields.
And it isn't to predict exactly how Google's AI systems will rank products.
It's to build a richer source of structured product knowledge for a search experience that is becoming more conversational.
Don't brainstorm questions. Mine them.
Make every product understandable. Make your top products discoverable.
Because the future of feed optimization isn't just telling Google what your product is.
It's helping Google understand when your product belongs in the conversation.
