Over the past few years, commerce leaders have grown accustomed to bold claims about AI. Conversational shopping will replace search. Websites will become irrelevant. Agents will do the buying for us.
Most of those claims have been directionally right and operationally vague.
At NRF this year, Google introduced the Universal Commerce Protocol (UCP). The announcement wasn’t a dramatic consumer-facing product launch with splashy user demos or an explicit mandate that all retailers must adopt it. Instead, UCP was presented as an open protocol intended to support what Google describes as agentic commerce, enabling AI assistants to interact with commerce systems through standardized APIs. In that sense, it landed as infrastructure rather than an event.
The announcement has triggered a range of responses across the industry. On the supportive side, major retailers and platforms like Shopify, Etsy, Wayfair, Target, and Walmart are listed as collaborators or early backers of the protocol, and media coverage highlights this coalition as a significant signal that multiple players see value in a shared standard for AI-mediated shopping.
Others were more cautious, raising concerns that protocols like UCP could further reduce the role of traditional websites and organic traffic, shifting even more control toward AI platforms. Google’s John Mueller addressed some of these concerns publicly, arguing that UCP is not intended to replace websites or eliminate SEO, but acknowledged that it represents another step in how commerce interactions may evolve.
To move past the reactions, let’s step back and look at the Universal Commerce Protocol on its own terms: what it is, how it works, and what it realistically changes for modern commerce organizations.
What Is the Universal Commerce Protocol?
The Universal Commerce Protocol is an open standard designed to allow AI clients to transact directly with commerce systems.
Announced by Google and backed by partners including Shopify, Stripe, other technology providers, and several large retailers, UCP defines a common way for AI agents to request and receive information such as product availability, pricing, promotions, loyalty rules, and authentication details, and then complete a transaction securely. Its focus is less on introducing new capabilities and more on standardizing how existing commerce logic is exposed across a fragmented ecosystem.
In practice, UCP acts as a shared language between consumer-facing platforms, businesses, payment providers, and identity providers. Rather than relying on custom integrations for each new surface, businesses can expose inventory and retail logic in a standardized form while remaining the merchant of record and retaining control over fulfillment and financial liability. Payment and credential providers handle sensitive data exchange using established security patterns, reducing compliance risk for other parties.
For the user, this means an AI assistant like Gemini can move beyond recommending products and into executing purchases without sending the user to a traditional ecommerce site. Discovery, eligibility checks, checkout, and payment all happen within the AI interface, while the transaction itself is still processed by the retailer and its chosen payment partners.
This idea is not new. UCP closely resembles theAgentic Commerce Protocol (ACP) developed by OpenAI and Stripe, which has already been adopted by some retailers and was used for live transactions during the most recent peak shopping period. Both reflect a broader shift toward agent-mediated commerce, where software acts on behalf of the user.
What makes UCP notable is not the concept, but the distribution. Gemini is embedded across Google Search, Android, and Chrome, giving the protocol a plausible path to reach consumers at scale. Understanding UCP at a high level is one thing; seeing how it would be adopted in practice is where its implications become clearer.
How Adoption Would Work in Practice
Despite some of the rhetoric, UCP does not replace a retailer’s commerce stack.
The flow is straightforward. An AI client initiates a request on behalf of a user. That request is sent via UCP to the retailer’s environment. The retailer responds through APIs or through a Model Context Protocol (MCP) host that exposes the necessary data and business logic.
The retailer still owns the systems of record. Product catalogs, pricing engines, inventory systems, promotions, loyalty programs, and authentication all remain on the retailer’s side.
What changes is the interaction layer. The experience happens inside the AI client, not on the retailer’s website.
Search and product discovery still have a role to play. At the same time, there is a real risk that AI clients like Gemini will increasingly augment or override merchant responses with their own AI-generated outputs, reducing the visible impact of third-party relevance and experience layers.
Once commerce interactions move into AI clients, an obvious question follows. If discovery, comparison, and checkout happen elsewhere, what role does the ecommerce site itself continue to play?
Does This Mean Ecommerce Sites are Going Away?
In theory, UCP points in that direction. In practice, the change is more gradual and uneven.
Claims about the “end” of ecommerce sites echo earlier predictions about the end of physical retail or brand websites. Those changes rarely eliminated existing channels. They added new ones. Physical retail adapted, and ecommerce sites are likely to do the same.
Conversational and agentic AI should be understood as another discovery channel. Some customers will research and buy through ChatGPT, Perplexity, or Gemini, and that is increasingly normal. Shoppers already move fluidly across platforms, and AI assistants are becoming part of that mix.
As AI-driven commerce grows, ecommerce sites may act less as destinations and more as services. Discovery, comparison, and even checkout can happen inside AI interfaces, with sites operating behind the scenes through APIs.
This has consequences. Traffic to owned channels may decline, and some traditional UX investments may matter less. But the larger issue is experience ownership. The retailers that perform best will participate in new AI channels while continuing to invest in strong search and product discovery on their own sites. Protocols enable access; owned experiences remain where brands differentiate and retain control.
What Happens to UX When the Experience Lives in AI?
UX does not disappear, but ownership shifts.
In a protocol-driven model, the experience is shaped by three things: the AI client, the data the brand exposes, and the rules defined by the protocol. Brands no longer control page layout, navigation, or presentation. They control the structure, completeness, and correctness of their data.
Google’s UCP announcement included an important nuance: business agents, which allow retailers to respond to customer queries in their own brand voice within Google surfaces. This gives brands a way to participate more directly in AI-mediated experiences, rather than being represented solely through generic summaries or third-party interpretation.
Even so, control remains limited. The AI client still owns the interaction framework, determines when and how brand agents are invoked, and governs ranking, presentation, and flow. Behavioral data and usage signals continue to accumulate primarily with the platform, creating learning advantages outside the brand’s direct reach.
For many retailers, this feels familiar. Google has long been an intermediary between brands and consumers. UCP extends that role from discovery into transaction. It is this loss of control, more than the technology itself, that explains why some industry responses have been cautious rather than celebratory.
Why the Industry Response Is Cautious
Two points stand out when looking at who is participating and who is not.
First, Amazon is notably absent. The company that controls a large share of US ecommerce Gross Merchandise Value has little incentive to expose its commerce capabilities through an open protocol controlled by someone else. Amazon has historically favored closed ecosystems and end-to-end control, and there is no sign that this strategy is changing.
Second, merchant sentiment is mixed. Shopify’s involvement is significant because it exposes a vast number of catalogs to AI channels, but it also highlights a growing concern: in agentic commerce models, merchants risk being downgraded to data providers. Product catalogs, pricing, availability, and fulfillment become the inputs that power AI experiences, while the AI platform owns the interface and the transaction.
What merchants get back is less clear. In many AI-mediated journeys, brands lose access to intent signals, rejection reasons, competitor context, feature importance, and conversational history. Those insights stay with the platform, limiting a merchant’s ability to refine discovery, merchandising, and personalization over time.
According to Forrester’s January 2026 analysis of agentic commerce, the pace of adoption is being shaped less by protocol availability and more by trust and execution. While Google and OpenAI are racing for platform advantage with UCP and ACP, Forrester’s research shows that many consumers still express hesitation about letting AI agents act on their behalf, particularly when it comes to payments, personal data, and error resolution.
In that context, agentic commerce is less about replacing existing discovery systems and more about raising expectations everywhere. As new AI-driven channels emerge, brands will be judged on how clearly, accurately, and consistently their products are represented — whether discovery happens in an AI assistant or on their own site.
Should Brands Move Everything to Google?
The short answer is no.
A consistent message from early adopters and industry leaders is to adopt emerging protocols like UCP without locking into any single platform. UCP lowers friction and expands reach, but it also introduces long-term strategic tradeoffs around control, data, and economics.
There is also a more immediate concern that is hard to ignore. Over the past decade, many retailers have successfully monetized their owned digital experiences through retail media. Search results, category pages, and recommendation zones are not just conversion surfaces; they are high-margin advertising assets powered by first-party behavioral data.
If discovery and checkout move into AI clients, retailers risk losing more than traffic. They risk losing control over how products are ranked and presented, access to the intent and behavioral data that powers retail media, and the analytics used to optimize merchandising and supplier relationships. The most valuable surfaces shift from owned channels to platforms controlled by someone else.
Early examples suggest this shift may not be economically neutral. In the emerging ChatGPT–Shopify checkout model, merchants are expected to pay a transaction fee of around 4% for sales completed through the AI interface, on top of existing payment processing costs. That mirrors a pattern retailers already understand from Google Shopping and marketplace ads: access comes with tolls, and visibility increasingly depends on platform economics.
Most retailers recognize this tradeoff, which is why adoption is likely to be incremental rather than wholesale. Framing UCP as an all-or-nothing decision misses the point. Its significance does not depend on universal adoption, but on how thoughtfully brands balance participation in new AI channels with continued investment in owned discovery and experience,
Why UCP Matters Even If It Is Not “The Future”
The most immediate impact of UCP has less to do with AI agents and more to do with data exposure.
Protocols like UCP remove the protective layer of a carefully designed UI. There is no page to compensate for missing attributes or inconsistent taxonomy. Weak data becomes visible very quickly.
This question is already common in enterprise conversations: “Is my data good enough for your AI?”
UCP makes that question unavoidable.
Retailers like Ecolab and US Foods have already experienced this reality. They discovered that digital transformation and data cleanup often have to happen at the same time. Waiting for perfect data is not realistic, but ignoring data quality is no longer an option.
As AI agents become more common entry points, data hygiene, completeness, and structure become table stakes. This shift in exposure and expectations also changes where value accrues in the technology stack, increasing the importance of strong search and product discovery foundations regardless of where the interaction begins.
Where Platforms Like Coveo Still Differentiate
Protocols like UCP and ACP signal a clear shift: conversational commerce is becoming an expectation, not an experiment. As shoppers grow used to AI-driven discovery elsewhere, they will expect similar experiences when shopping on a retailer’s own site.
That expectation creates an opportunity. AI agents operating outside owned channels are optimized for efficiency, but they are limited in context and control. They favor popular signals and standardized flows, and they abstract away brand-specific nuance, complex assortments, and long-tail relevance.
Platforms like Coveo help retailers meet rising expectations for conversational discovery on their own terms. By bringing intelligent search and product discovery directly into owned experiences, retailers can deliver the speed and relevance shoppers expect while retaining control over data, merchandising, and economics.
Just as importantly, this model is more profitable. Unlike agentic platforms, Coveo does not sit between the retailer and the transaction or take a cut of conversions. Retailers can use conversational discovery to grow revenue and loyalty without trading margin for access.
A Note on B2B
It is also important to be clear about scope. UCP is, for now, largely a B2C phenomenon.
The B2B commerce market is massive and growing rapidly, but it operates very differently from consumer retail. According to recent figures, globally, B2B ecommerce is already worth more than five times the size of B2C, estimated to be worth $32.1 trillion in 2025, with projections reaching $62.2 trillion by 2030.
Business-to-business transactions involve contract pricing, entitlements, regulatory constraints, and confidential supplier relationships — data and logic that are rarely exposed in a simple API feed. A consumer buying a jacket is not comparable to a hospital sourcing pharmaceuticals under negotiated terms, or a manufacturer managing supplier contracts and bulk pricing tiers. B2B ecommerce involves large order volumes, structured pricing, compliance considerations, and bespoke fulfillment rules.
In many B2B contexts, interactions still require negotiation, manual review, credit terms, and fulfillment coordination that are not captured by today’s agentic commerce standards. This limits how universal any current commerce protocol can be.
Final Thoughts
UCP is not the end of ecommerce, and it is not the future on its own. It is a signal of where commerce interfaces are moving and how competition for control is shifting.
As conversational and agent-driven experiences become more common, they will shape customer expectations everywhere — including on retailers’ own sites. The question is not whether these new interfaces will exist, but how brands participate without giving up control over discovery, data, and economics.
The practical response is measured experimentation. Adopt new protocols thoughtfully, invest in data readiness, and ensure owned experiences evolve to meet rising expectations for intelligence and ease.
Key Takeaways
- UCP and ACP signal the rise of agentic and conversational commerce, but they are infrastructure, not strategy.
- Conversational experiences will increasingly be expected on retailers’ own sites, not just in AI platforms.
- Agentic commerce introduces new reach, but also shifts control over discovery, data, and monetization surfaces.
- Early economic experiments, including transaction fees and ads, raise questions about future value capture rather than settling them.
- Strong search, product discovery, and data foundations remain critical across all channels.
- Adoption is likely to be incremental, shaped by trust, execution, and economics rather than protocol availability.