Industry leaders who have been in e-commerce much, much longer than me are saying this is the greatest impact yet to agentic commerce.

Scott Wingo said in his latest episode of the Retailgentic Podcast that he has been in the business for 30 years and has "never seen anything like what we are seeing right now". Google Cloud VP, Carrie Tharp, also echoed this sentiment, calling the shift to agentic commerce the industry's "most significant transformation yet".

Let's go Agentic Commerce!!! Super exciting to see this happening!

And while I'm very excited about what happens next, that level of conviction actually makes me more curious than anything...

Universal Commerce Protocol - UCP

At the center of this launch is Universal Commerce Protocol, or UCP. It's being positioned not as another API, but as a "shared language", a protocol that allows AI agents to move from passive browsing to autonomous action.

The reason that this is such a big deal is because the announcement is coming from Google, and the capabilities, along with the partnerships show serious investment in the infrastructure of agentic commerce, and the retailers and marketplaces that will be empowered through this protocol.

At a practical level, UCP breaks commerce down into a set of modular capabilities that agents and businesses can understand and work with dynamically. Instead of hard-coding every integration, businesses expose what they can support, things like product discovery, cart, checkout, and payments, and agents discover those capabilities in real time.

UCP Diagram - Detailed architecture showing how agents discover capabilities in real time

Those capabilities act like building blocks. Checkout is a capability. Product discovery is a capability. Discounts, delivery options, or loyalty perks can be layered on as extensions. An agent doesn't need to know how a merchant is implemented internally; it just needs to know what capabilities are available and how to invoke them.

Payments are treated the same way. UCP separates how a consumer pays from who processes the payment, which allows agents to work across a wide range of existing payment providers without custom logic for each one. And because UCP supports multiple ways for systems to talk to each other, APIs, MCP, and agent-to-agent (A2A) protocols, businesses can adopt it without ripping out their current stack.

The end result is a system where agents can discover what a business can do, understand how to transact, and complete purchases without every merchant and surface having to build one-off integrations for each other.

Just a call back to a few months ago in a post I made, I believed that MCP was not enough, and it certainly looks like that is true with OpenAI's ACP and now the UCP protocol being created.

Google notes that e-commerce has always been defined by "N x N" integrations, where every merchant, every platform, and every surface had to build custom connections. Google developers explain that this previous model breaks down once the interface itself becomes an agent.

UCP attempts to collapse that complexity into a single abstraction layer.

And with more than 20 partners developing alongside Google and Shopify across the ecosystem, it's impossible to not see the impact.

And with mass adoption, if the default shopping surface becomes an agent rather than a page or a search result, what does it actually mean to "win" demand? Previously there were some predictions on the next e-commerce wave not having storefronts... will this signal some retailers to try completely headless experiences?

From Keywords to Content and Context

One of the most important shifts here following UCP and also ACP is how discovery works, and how important it is in the agentic commerce channel.

Agents don't operate on keywords... they need content and context.

If I go to ChatGPT, Gemini, or Perplexity, and ask for help finding "red running shoes under $100", what products will be surfaced to me? If I specify that I need them specifically for trail running or sprinting, then how do the surfaced products change? Add in a brand, and suddenly you're able to see what that discoverability looks like across the LLMs.

This is the content and context that are needed for these agents, and optimization for both might be the most important win that I believe retailers/marketplaces can focus on right now.

The New Measurement: Prompt to Product

Ranking for keywords is then no longer sufficient for agentic discovery. What matters now is understanding which prompts surface which products across different buying agents.

Instead of asking, "do we rank for this term?" you're asking, "when someone asks an agent for this exact thing, what shows up, and why?". This introduces a new measurement problem. Prompt-to-product surfacing becomes the primitive, not impressions or clicks.

Relevance to #GEO & #ACO?

In a previous post I wrote, I said that the conversation around SEO/GEO/LLMEO felt largely speculative. Lots of incredible ideas, but little measurement, especially given that we're all working with foundational models that are effectively black boxes.

After finally seeing work that showed seller-side content actually moving outcomes inside a generative shopping experience, it became much more real...

And I don't think that's a separate conversation from what's happening with UCP. I think it's the same conversation, just at a different layer.

UCP shapes how agents transact, while GEO influences what they surface and prioritize before any transaction occurs.

If UCP becomes the shared language that allows agents to discover and transact, then GEO becomes one of the first real seller-side levers inside that system. It's how sellers influence what agents see, how they interpret value, and what they recommend, before any checkout even happens.

Put differently, when I said before that seller agents already exist and are just prompt-driven for now, this is what I meant.

So these are the questions that I would be asking if I was a brand/retailer/marketplace around agentic optimization are:

Shopify's Agentic Storefronts and the Open Catalog

One additional announcement that really stood out to me is Shopify's move to open its catalog infrastructure beyond Shopify-hosted storefronts through its new "Agentic plan".

With Agentic storefronts, merchants can now manage their product data once and have it surfaced across multiple AI channels, including Microsoft Copilot and Google AI Mode. That alone feels important, but what's more interesting is what this represents conceptually.

If agents are becoming the discovery layer, then the catalog starts to look less like a backend system and more like a seller-side interface to agentic demand.

This is where I keep thinking back to earlier posts on #seller #agents! Even if we don't call this a "seller agent" yet, it feels spot on. It's a structured, centralized way for sellers to express product information, constraints, and context in a form that agents can actually reason over.

Shopify's CEO has described this as enabling "serendipity", where an agent finds the right product for a customer who might never have searched for it directly. If the catalog optimized by LLMs becomes the source of truth for agents, distribution starts to look very different.

Instead of optimizing only for human shoppers on a single storefront, sellers can begin optimizing for how their products are interpreted and surfaced across agentic channels.

That feels like a meaningful step toward the seller-side optimization many of us have been expecting, and it's hard to imagine this being the last iteration we see.

Agentic Commerce is the word for 2026

If this doesn't cement agentic commerce as here to stay, it's hard to imagine what would. This is so much further past a trend, as it completely changes how discovery works, how decisions are made, and where trust is built. If protocols like UCP and ACP see broad adoption, "I bought this from ChatGPT" may soon replace "I bought this on Amazon."

Sources

*posted originally on LinkedIn - link*