Agentic Shopping Has Arrived

The Agent Is Now Shopping for You

Something significant happened in the retail landscape in 2025 that most marketers have not yet fully processed.

AI agents — autonomous software systems that can browse websites, compare products, make decisions, and execute purchases on behalf of users — became genuinely useful and accessible to mainstream consumers. OpenAI's Operator, Anthropic's Computer Use, and Google's Project Mariner demonstrated that the agentic shopping era is not a future scenario. It is happening now.

The implications for physical and digital retail are profound.

What Agentic Shopping Looks Like

Traditional e-commerce assumes a human at the keyboard. The shopper browses, compares, reads reviews, and clicks Buy. Every step of this journey has been optimised for human attention — product photography, UX flows, copywriting, promotional messaging.

Agentic shopping removes the human from most of that journey.

A user instructs their AI agent: "Find me a pair of white running shoes under €120, men's size 43, available for next-day delivery in Barcelona." The agent then:

1. Searches across multiple retailers and marketplaces

2. Reads product pages, specifications, and availability data

3. Cross-references reviews and ratings

4. Compares prices across vendors

5. Presents a recommendation — or, with permissions granted, completes the purchase directly

The agent doesn't see your product photography. It reads your product data. It doesn't respond to emotional copywriting. It processes structured attributes. It doesn't visit your homepage. It queries your product catalogue.

The Data Layer Is the New Storefront

For retailers operating in the agentic shopping era, the critical asset is no longer the website as a visual experience. It is the data layer underneath that experience.

Agents need:

Structured, attribute-rich product data — Every product needs complete, standardised attributes: size, colour, material, availability, price, dimensions, and any domain-specific attributes relevant to the product category. Incomplete data means the agent skips your product.

Machine-readable accessibility — Product data must be accessible to AI systems without JavaScript execution. Server-rendered HTML with Schema.org Product markup is the baseline. MCP (Model Context Protocol) server endpoints are rapidly becoming the standard for real-time agent access.

Accurate inventory signals — An agent that recommends a product that is then shown as out-of-stock at checkout learns quickly not to trust that retailer's data. Real-time inventory accuracy is an agent trust signal.

Clear purchase pathways — The agent needs to understand how to complete a purchase without human assistance. Clear, well-structured checkout flows with minimal friction are essential.

What This Means for Physical Retail

Physical retailers face an additional layer of complexity. Their inventory exists in physical stores, not just online warehouses. Availability varies by location. Prices may differ between channels. Store hours matter.

Shopping centres face an even more acute version of this challenge. Their tenant data — which stores sell what products at what prices — is often poorly structured, inconsistently maintained, and entirely inaccessible to AI systems.

GoNow Luma was designed to solve precisely this problem. By creating a structured, real-time data layer for shopping centre tenants — complete with Schema.org markup, MCP server access, and vector-embedded product search — we make physical retail legible to AI agents for the first time.

Getting Ready for Agentic Commerce

For any retail brand or shopping centre, the readiness checklist for agentic commerce looks like this:

• Product data quality: Are all products attribute-complete? Are specifications standardised and accurate?
• Technical accessibility: Is product data available as server-rendered HTML with Schema.org markup? Is there an API or MCP endpoint for agent access?
• Inventory accuracy: Is real-time inventory availability exposed in a machine-readable format?
• AI crawler access: Does robots.txt welcome GPTBot, ClaudeBot, and other AI crawlers?
• Structured search capability: Can an AI agent search your product catalogue by attribute without human assistance?

The brands that answer yes to these questions today are building a compounding advantage. The brands that postpone are watching their products disappear from the most rapidly growing consumer discovery channel in a generation.

About the author: Jason Leven is CEO and Co-Founder of GoNow Productions, a GEO and AI digital agency based in Barcelona. He has 25+ years of experience in software development, digital search, and SEO across 21 countries.

GoNow Productions produces this content and offers commercial services in AI search optimisation for retail.