Google'sAIBusinessAgentIsLiveinMerchantCentre.YourFeedJustBecameaSalesChannel.

Google has launched an AI Business Agent inside Merchant Centre in the US. This is direct from one of our clients' accounts.

If history tells us anything, the UK will not be far behind.

This is not just another chatbot.

It sits directly on top of your product feed. That means it can:

• Answer product questions using your structured data
• Surface items on sale or "best deals"
• Highlight new arrivals or best-rated products
• Guide shoppers before they ever reach your site

In other words, your feed is no longer just powering ads. It is powering conversations.

What the Business Agent actually does

The AI Business Agent is a conversational layer that Google is embedding directly into the shopping experience. It draws entirely from your Merchant Centre data - your product titles, descriptions, attributes, pricing, promotions, and availability - to answer shopper queries in natural language.

Think of it as a shop assistant that knows your entire catalogue. But unlike a shop assistant, its knowledge is limited to exactly what you've told Google. Every product title, every description, every custom label, every promotion - that's its training data. Nothing more.

Inside Merchant Centre, brands can customise the agent with a welcome message, conversation starters, brand colours, and a handoff message for when the AI can't answer. It's designed to feel like a branded experience, not a generic Google chatbot.

But here's the critical point that most advertisers will miss:

The quality of the conversation is entirely determined by the quality of your feed.

What changes for UK merchants

This feature is currently live in the US. But Google's rollout pattern is predictable: US first, then English-speaking markets, then everywhere else. UK merchants should be planning for this now, not reacting when it arrives.

Here's what shifts:

Feed quality becomes a competitive advantage

If your titles are generic and your attributes are thin, the agent has nothing useful to work with. It can't recommend a product it doesn't understand. It can't highlight a benefit that isn't in the data. It can't answer a question about sizing if your size attributes are missing.

If your data is structured properly - enriched titles, complete attributes, accurate categorisation, detailed descriptions - you win more visibility inside conversational queries. Your products become the ones the AI recommends.

This is the same principle we've been advocating for years with Google Shopping and Performance Max. The algorithm can only work with the data you give it. The AI Business Agent just makes that dependency more visible and more consequential.

Promotions strategy becomes more exposed

If you lean heavily on discounting, the AI will surface that narrative. "What's on sale?" becomes a natural conversational query, and the agent will happily tell shoppers about every markdown in your catalogue.

For brands that protect margin at SKU level, this is manageable. You can control what gets pushed and what doesn't through your feed strategy, custom labels, and promotion configuration.

For brands running blanket promotions with no margin logic, the agent will expose the full extent of your discounting strategy to every shopper who asks. That's a commercial risk most performance teams haven't considered.

Intent signals get stronger

Conversational queries tied to products are high intent. Someone asking "What's the best waterproof jacket under £150?" is further down the funnel than someone typing "waterproof jacket" into a search bar.

That conversational data will inevitably feed into Smart Bidding and PMax optimisation. Google will use the signals from these interactions - what shoppers ask about, what they respond to, what leads to a click - to refine how it serves your products.

Brands with better feed data will generate better conversational interactions, which will generate better signals, which will improve bidding efficiency. It's a compounding advantage.

The gap this will widen

This is where it gets uncomfortable for a lot of ecommerce brands.

Brands treating Merchant Centre as admin will fall behind. Brands treating the feed as infrastructure will compound.

If you are running thousands of SKUs inside one broad PMax campaign with no margin logic, no custom labels, no SKU-level profitability driving bidding - this will widen the gap between you and brands that do.

If you have clean taxonomy, enriched attributes, and SKU-level profitability driving your bidding strategy - this is an opportunity. The AI Business Agent will serve your products more effectively, more often, and to higher-intent shoppers.

The divide isn't about budget. It's about infrastructure.

What this means for feed architecture

The AI Business Agent changes the role of the product feed fundamentally. It's no longer just a data source for ad targeting and product listings. It's now a data source for a conversational sales experience.

That means the things that were "nice to have" in your feed are becoming essential:

• Detailed product descriptions - The agent needs natural language to work with. Thin, keyword-stuffed descriptions that were written for old-school SEO will produce terrible conversational responses.
• Complete attribute coverage - Size, colour, material, style, occasion, season. Every empty attribute field is a question the agent can't answer.
• Accurate pricing and availability - The agent will surface pricing and stock status. If your data is stale, the conversation will be misleading.
• Structured promotions - Sale prices, promotional messaging, and deal annotations need to be configured properly in Merchant Centre, not just on your website.
• Product ratings and reviews - If you have product ratings configured, the agent can use them to recommend "best-rated" products. If you don't, your competitors' products get recommended instead.
• Custom labels for commercial logic - Custom labels that reflect margin tiers, inventory velocity, and commercial priority will determine which products the agent surfaces in response to value-oriented queries.

This is not a cosmetic update. It's a structural shift in how Google uses your product data. And it requires a structural response.

Why this reinforces the POAS approach

For brands running a POAS methodology, this development is an accelerator.

If you already know which products are genuinely profitable - not just revenue-generating - you can configure your feed to ensure the AI Business Agent surfaces the right products. High-margin products with complete data will be recommended more often and more effectively.

Conversely, if you're still running on ROAS-based targets with no margin data in your feed, the agent will happily recommend your lowest-margin products to every shopper who asks for a deal. It doesn't know they're unprofitable. You haven't told it.

The feed is no longer just an input into auction dynamics. It's an input into a sales conversation. And like any sales conversation, you need to control the narrative.

What to do now

The UK rollout timeline is unknown, but the preparation is straightforward:

1. Audit your feed completeness. Check attribute coverage across your entire catalogue. Identify gaps in descriptions, categories, and product attributes that would limit the agent's ability to answer questions.
2. Review your product descriptions. Rewrite them for conversational use, not just keyword density. The AI needs natural language that accurately describes benefits, features, and use cases.
3. Map your custom labels to commercial logic. Ensure your custom labels reflect margin tiers, inventory priority, and commercial strategy - not just arbitrary groupings.
4. Assess your promotions strategy. Understand what the AI will surface when shoppers ask about deals. If that narrative doesn't serve your margin targets, adjust your promotion configuration.
5. Configure product ratings. If you have reviews, make sure they're flowing into Merchant Centre. This is an underutilised data source that the agent will leverage.
6. Verify your Merchant Centre health. Account-level issues, policy violations, and data quality warnings will affect how the agent performs. Clean accounts will generate better conversational experiences.

Worth reviewing your feed architecture now rather than when CPCs quietly rise and you cannot explain why.