AI Overviews and Conversational Attributes: Get Your Products into Google's AI Answers

We no longer search for queries like "running shoes". Instead, we ask "what shoes for running in the rain under $100" or "what to wear for running when I have flat feet". These long queries are processed by AI Overviews, which is powered by the Gemini model. The result is a list of recommended products and links.

If your product feed only contains a basic product name like title: "Adidas Adizero, 42 blue", the AI has nothing to match against the query. Your product won't show up in the AI answer. This article shows how to adjust your feed so the AI can read your products correctly and recommend them.

What AI Overviews is

AI Overviews are the panels Google has started inserting above traditional search results. Instead of the usual handful of blue links, we now see a short text summary with links to sources.

For shopping queries, AI Overviews often show a product carousel (this feature is currently rolling out in the Czech Republic). It contains three to six products from different online stores. For each product, it shows a photo, a price, and a short explanation of why it was selected.

The main difference from traditional search is clear: AI Overviews answers the question directly instead of just pointing to the best website. If your product doesn't have correctly structured data, users won't see it. This was already true for SEO, and now it applies to PPC as well.

How AI Overviews picks products

Here's the key difference: AI Overviews doesn't select products by keywords. It uses conversational matching.

Traditional Google Shopping relied on direct keyword matching for a long time. Gemini, however, works differently. It analyzes text contextually, breaking the query into semantic units to understand the user's intent.

Consider the query: "what shoes for running in the rain under $100 for flat feet". The system internally links "rain" with a wet surface and "flat feet" with a need for stability (pronation). If you don't have these nuances described in your feed in a structured way, the AI model simply won't recommend your product for such a specific query. For example, a generic description like "Comfortable running shoes for every runner" won't get you anywhere.

Google gets contextual information (e.g., about running on asphalt) in three ways:

1. Via the product_highlight attribute (key benefits). This attribute is for short, punchy bullet points about product properties. The AI model parses these with high priority, as it knows they contain important technical specifications.

   Example in an XML feed: <g:product_highlight>Suitable for running on asphalt and hard surfaces</g:product_highlight>.

2. Via semantic extraction from the product description and title. Gemini works differently than traditional search engines; it looks for deeper meaning. If your description contains the sentence "The sole absorbs shocks on city streets," the AI can infer a use case for asphalt surfaces.

3. Via Conversational attributes: 6 official fields in GMC

Google names six attributes in its documentation, labeling them conversational attributes. They are meant to help AI systems better understand the specific details of products. As a primary use case, Google mentions AI Mode and traditional search. AI Overviews shares the same infrastructure, so the investment pays off here as well.

1. question_and_answer: FAQs in the feed

These are question-and-answer pairs. AI Mode uses this attribute for queries like "does the product have this feature?" or "does it fit a specific model?". Haven't we already seen something similar in SEO?

"Does it fit a 15-inch MacBook Pro?": "Yes, internal dimensions 36 × 25 cm fit a 15-inch MacBook Pro."
"Does it have a laptop sleeve?": "Yes, a padded sleeve for devices up to 16 inches."

2. document_link: links to PDF documents

This is for manuals, technical sheets, or product brochures. If a user enters a specific technical query, the AI can offer a direct link to the document.

3. related_product: relationships between products

This attribute defines accessories or replacement parts. It contains parameters like relationship type and identifier. This allows the AI to recommend compatible goods, for example, linking a coffee machine directly to the correct filter.

4. item_group_title: a shared title for variants

When a product has many colors and sizes, this field unifies them. The AI then groups the variants under a single product in its answer instead of showing them separately.

5. variant_option: structured variant properties

This enables precise filtering by parameters like shoe width or material, allowing a query for a specific variant to match exactly.

6. popularity_rank: popularity score

This is a numeric value from 0 to 100 that serves as a useful signal for ordering products in AI answers.

Supporting attributes: AI reads them too

These fields are not officially labeled as conversational; they are general enrichment attributes. However, the AI reads them to improve query matching.

material: State this explicitly in a separate field. It's not enough to have it hidden in the description text.

product_highlight: List three to five clear benefits. Each bullet point should be 50–70 characters. Longer text gets truncated on mobile and is harder to parse. A concrete bullet point has a better chance of being featured.

age_group, gender, size: Send these as separate, structured demographic fields. Don't just include them in the product name.

When you add material, key benefits, and stability technology to the feed, your chances of appearing in an AI answer increase significantly.

Platform dependency vs feed control

With traditional SEO, you write articles and gather backlinks. Google's crawler periodically crawls your site and ranks you in the results. Your main levers are your site's text and domain authority (apologies to my SEO colleagues for the simplification 😀).

With AI Overviews, it's different for shopping queries. When the product carousel appears, Google isn't looking at your blog. It's searching the massive Google Shopping Graph database. The easiest, most controllable way into this database is through your product feed in Google Merchant Center (GMC).

How to fill attributes for thousands of products

Populating attributes by hand is quick for a few items, but it's not efficient for a large catalog. This is where it makes sense to use AI. It can help you bulk-generate questions and answers from technical specifications or correctly establish relationships between products from categories. To change the values in the feed, you don't need a developer; a supplemental feed in Google Merchant Center is all you need.

Concrete examples (before / after)

Product 1: men's winter jacket

Before:

• Title: Jacket M-501 blue
• Description: Men's winter jacket, blue, size M.

After:

• Title: Alpine Pro Carlos men's waterproof winter jacket, blue M
• Description: Waterproof winter jacket with a 10,000 mm membrane. Inner fleece lining, removable fur hood trim, reflective elements. Suitable for winter hiking and skiing down to -15 °C.
• material: polyester
• product_highlight: 10,000 mm membrane
• question_and_answer: "How cold can it handle?": "The lining keeps you warm down to -15 °C."

AI Overviews is more likely to feature the edited product. Before the edit, the system had no reason to include it. The Q&A also addresses real customer questions.

Product 2: hydrating cream

Before:

• Title: Cream 50ml
• Description: Hydrating cream.

After:

• Title: La Roche-Posay Toleriane hydrating day cream for sensitive skin, 50 ml
• Description: A hydrating day cream for sensitive and reactive skin. Fragrance-free and hypoallergenic. Suitable for use after dermatological procedures.

Product 3: kids' cycling helmet

Before:

• Title: Kids' helmet

After:

• Title: Bell Sidetrack II MIPS kids' cycling helmet, blue M (52-56 cm)
• Description: A kids' cycling helmet with MIPS technology to protect against rotational forces during a fall. Features an adjustable sizing system. Suitable for children ages 5 to 10.

6 quick actions for your feed

1. Add three Q&A pairs to your bestselling products. Use common queries from customer support.

2. Fill in the material attribute for your top products. You can often map this from existing data.

3. Add three to five clear benefits using product_highlight. Avoid generic phrases.

4. Add variant properties (variant_option) for products with multiple configurations. The AI will then group the items correctly.

5. Split age, gender, and size into separate fields.

6. Ensure your google_product_category assignments are as specific as possible.

Watch out: AI Overviews have their flaws

The technology is live, but it's still evolving. You may see price inaccuracies or incorrect category assignments. Product positions in the AI carousels also fluctuate significantly. But the investment in data won't go to waste. The same information is used by AI Max for Shopping, traditional Shopping, Google Lens visual search, and autonomous agents in Agentic Commerce.

Measurement in Google Ads and SEO

In Google Search Console, you'll find a filter for AI Overviews, but it has limitations. If the AI cites your product without a user click, the impression often isn't counted.

In Google Ads, measurement is currently limited by the technology. Click and impression data from AI carousels are blended with standard Shopping and PMax campaign data in reports. Isolating the exact ROAS generated exclusively by AI Overviews isn't possible right now.

As PPC specialists, we shouldn't chase one isolated number in the account. It's more important to watch the overall trend. If you see an overall increase in CTR and conversion rate after enriching your product data, that's a clear signal the optimization is working.

Summary

1. AI Overviews select products based on conversational matching, not keywords.
2. Google defines six official conversational attributes to help it understand product details.
3. Other fields, like material and product_highlight, also help.
4. Start by focusing on your bestselling products.
5. Good data also helps with PMax campaigns and Google Lens.
6. The impact of AI carousels is still difficult to measure precisely, so monitor overall trends.

In 2026, AI search is playing a major role. Google is enabling it for more and more users. Without conversational attributes, your products will be less visible in these systems.