Gap Adopts UCP: How Brands Can Optimise for Agentic Shopping in Gemini

Gap has become the first major fashion retailer to launch Instant Checkout inside Google Gemini. Powered by Google’s Universal Commerce Protocol, shoppers can now browse Gap’s catalogue, receive styling recommendations, and complete checkout entirely within Gemini. No redirect, no website, just conversation to cart to purchase.

As Gap’s CTO Sven Gerjets put it, the shift is no longer about keyword search, but about conversations. Customers are no longer typing “blue dress” into a search bar. They are asking, “What should I wear to a wedding?” or “What’s right for a job interview?” and expecting a complete answer, not a list of links.

Why this moment matters

This launch coincides with a broader set of updates from Google that make this experience viable at scale.

Universal Commerce Protocol now introduces persistent shopping carts, allowing customers to build and manage baskets directly inside Gemini. Brand catalogues provide agents with live product data, including variants, inventory, and pricing. Identity linking ensures shoppers retain loyalty benefits and account level perks, even within the AI interface.

Taken together, these updates move agentic commerce from concept to something that works at a consumer level.

The two shifts behind Gap’s move

Gap’s adoption reflects two underlying realities. Agentic commerce is no longer an early experiment, it is operational. At the same time, early adopters are gaining a meaningful advantage by shaping how they appear in this new discovery layer.

By integrating with UCP, Gap ensures its products are both purchasable and accurately represented at the exact moment customers express intent.

The new battleground: discovery inside Gemini

For other brands, the question is not whether this shift is happening, but how to respond.

Visibility becomes the first hurdle. If a product is not surfaced in the conversation, it cannot be purchased. UCP enables transaction, but it does not guarantee recommendation. Discovery becomes the competitive layer, where brands are selected or ignored by the agent itself.

How brands can optimise for discoverability in Gemini

1. Content: make products legible to agents

Create comprehensive specifications for every SKU, but think beyond completeness towards precision. Agents do not interpret loosely written marketing copy well. They respond to explicit, structured information that maps cleanly to user intent.

Feature lists should separate what a product is from what it does. Benefit summaries should connect those features directly to outcomes. FAQs should not be generic support content, but designed around real conversational queries such as “Is this suitable for winter weddings?” or “Will this work for wide feet?”

Certifications, compliance details, and safety information are not edge cases. They are often decisive factors in agent recommendations, particularly in regulated or high consideration categories.

Use cases should be concrete and situational. “Smart casual office wear” is weaker than “suitable for a first day in a corporate office” or “appropriate for a summer outdoor wedding.”

The underlying principle is simple. If an agent cannot confidently answer a user’s question using your product data, it is less likely to recommend your product at all.

2. Context: define who the product is for

Agents do not just match keywords. They match intent. Context is what allows that match to happen accurately.

Defining target users should go beyond broad demographics. It should capture needs, constraints, and scenarios. A product might be for “commuters cycling in urban environments” or “guests attending formal evening events,” not just “men aged 25 to 40.”

Usage context should include timing, environment, and occasion. When is this product relevant, and when is it not? This includes seasonality, formality, and even cultural expectations.

Compatibility becomes critical in narrowing recommendations. This can include sizing systems, device compatibility, accessory pairings, or environmental constraints. The more explicitly these relationships are defined, the less guesswork the agent has to perform.

Equally important are limitations. Products that clearly state what they are not suitable for often perform better in agent systems, because they reduce the risk of a poor recommendation.

3. Authority: establish trust signals

Agents are selective. When multiple products meet the same criteria, they tend to favour sources they recognise as credible.

Authority is built through consistency. Your product data, brand messaging, and external presence should align across every surface where your brand appears. Discrepancies create uncertainty, which reduces the likelihood of recommendation.

Owned content plays a foundational role. Detailed product pages, buying guides, and supporting content help establish depth and expertise. Organisational schema reinforces this by making brand information machine readable and verifiable.

Earned presence is increasingly influential. Mentions in reputable publications, marketplaces, and structured data sources feed into the training and retrieval layers that agents rely on. If your brand is absent from these sources, it is less likely to be surfaced.

Monitoring becomes part of the process. Brands need to understand where they are being referenced, where they are missing, and how they are being represented across the ecosystem.

4. Structured data: the operational foundation

Structured data is what allows agents to reliably retrieve, compare, and transact on products. Without it, even strong content and context can become inaccessible.

Product schema should be complete and validated, but more importantly, consistent. The same product should not have conflicting attributes across your site, Merchant Center, and other platforms. Inconsistency introduces friction at retrieval and can lead to exclusion.

Product feeds should be treated as live infrastructure. Pricing, availability, and identifiers must be continuously updated. Delays or inaccuracies do not just affect user experience, they affect whether agents trust your data at all.

Identifiers are particularly important. Variant IDs, SKUs, and global identifiers need to be stable and correctly mapped, as they are the link between discovery and checkout.

As platforms expand the number of available attributes, depth becomes a differentiator. Brands that populate richer datasets give agents more signals to work with, increasing both visibility and relevance.

5. Measurement: rethink visibility

Traditional analytics were built for a world where users clicked links. Agentic commerce operates earlier in the journey, often before a click ever happens.

This requires a shift in what is measured. Brands need to understand how often they are appearing in AI generated responses, in what contexts, and for which types of queries. Presence alone is not enough. Positioning within those responses also matters.

Inclusion and exclusion become key signals. If competitors are consistently recommended in scenarios where your products are relevant, that indicates a gap in either data, authority, or context.

Visibility should also be tracked across platforms and regions, as agent behaviour can vary depending on the ecosystem and market.

Over time, agent visibility becomes a leading indicator of demand. It shapes consideration before traffic, and increasingly determines whether a customer ever reaches your site at all.

From discovery to checkout: Preparing for UCP

The implications are clear. Optimising for discoverability is no longer just a marketing exercise, it is a prerequisite for participation. But visibility alone is not enough. Once a brand is surfaced, it must be able to transact seamlessly within the same environment.

That is where Universal Commerce Protocol moves from theory to implementation. The shift from being recommended to being purchased happens through the infrastructure that sits behind these experiences. For brands, the next step is understanding how to plug into that layer, and how to structure their systems so that discovery, recommendation, and checkout operate as a single flow.

How to get started with the Universal Commerce Protocol

Google’s UCP implementation is already live across Gemini and AI Mode in Search, making it the most direct way to reach users in conversational shopping environments.

Requirements

To participate, brands need an active Google Merchant Center account, products that are eligible for checkout, and structured, accurate product data that can be reliably surfaced and transacted on.

Setup steps

The process is relatively straightforward. Set up or verify your Merchant Center account, upload and validate your product feed, complete Google’s merchant interest form, and follow the UCP integration guide. Once approved, your products can be discovered and purchased directly inside conversational experiences without redirecting to your site.

Structuring product data for UCP

Access to UCP is only one part of the equation. The quality and structure of your product data determines whether agents can understand, recommend, and transact on your products effectively.

The core model: product and variant

UCP is built around a clear hierarchy.

A product represents the overall catalogue item, including title, description, media, and a set of variants.

A variant is the purchasable unit. It reflects a specific configuration such as size or colour, and carries its own price, availability, and identifier. This identifier is critical, as it is the same one used at checkout.

Agents do not buy products. They buy variants.

What good product data looks like

At a minimum, every product should include a clear title, a detailed description, a defined price range, and a well structured set of variants. Media should be ordered by relevance, with the most representative image first, as agents will treat this as the default.

Variants should include precise titles such as “Blue / Large”, accurate pricing, availability status, and selected options. Where relevant, include SKU or barcode data to support identification across systems.

The emphasis throughout is on clarity and consistency. Ambiguous or incomplete data reduces the likelihood of accurate recommendation.

Pricing and availability

Pricing in UCP is explicit and structured. Every price includes both an amount and a currency code, supporting multi currency environments. Price ranges at the product level allow agents to understand variation across options.

Catalogue responses reflect current pricing and availability, but they are not transactional commitments. Checkout remains the authoritative step. This means data must be continuously updated and revalidated rather than cached or reused across sessions.

Context and personalisation

UCP introduces a flexible concept of context, which helps agents tailor results to user intent and location.

This can include signals such as country, region, postal code, language, and currency, as well as softer inputs like user intent, for example “looking for a gift under £50” or “something durable for outdoor use”.

These signals guide relevance rather than define it. Brands can use them to adjust pricing, availability, and recommendations, while still enforcing final rules at checkout.

Search, discovery, and retrieval

The catalogue layer supports several key behaviours that underpin agentic shopping.

Free text search allows agents to retrieve products based on natural language queries. Category and filter based browsing supports more structured exploration. Lookup enables direct retrieval of products or variants by identifier, which is especially important when moving from recommendation to checkout.

All of these rely on well structured data. If products are not clearly categorised, tagged, and described, they are less likely to be retrieved in the first place.

The link to checkout

Every catalogue interaction ultimately leads to checkout. The critical connection is the variant ID.

When an agent selects a product, it passes the corresponding variant ID into checkout as a line item. If this ID is inconsistent or missing, the transaction cannot proceed.

This makes identifier integrity one of the most important, and often overlooked, aspects of UCP readiness.

What this means in practice

Getting started with UCP is increasingly accessible, but success depends on more than integration. The brands that perform well will be those that treat their product data as infrastructure, not content.

In a world where commerce is mediated by agents, your catalogue is no longer just a backend system. It is the interface through which your products are understood, recommended, and ultimately purchased.

Summing Up

The shift to agentic commerce is already underway. Gap’s adoption of UCP is an early signal, but not an isolated one. As conversational interfaces become a primary layer of discovery, the brands that win will be those that are both visible and operational within these environments.

This is not just a new channel. It is a restructuring of how commerce works, from discovery through to transaction.

At Azoma, we help brands prepare for and win in agentic commerce.

From structuring product data and optimising for discovery, to implementing UCP & AMP, we work across the full stack of what it takes to be recommended and purchased inside AI systems.

➡️ If you are thinking about how your brand shows up in Gemini and beyond, get in touch with us today.