Universal Commerce Protocol: Why Google's UCP is turning the e-commerce landscape upside down

Google has made a strategic move with the Universal Commerce Protocol. Most people are overlooking the crucial point here. Here is my analysis after more than 10 hours of research.

On January 11, 2026, Sundar Pichai appeared in person at the NRF conference in New York. That doesn't happen often. When Google's CEO appears in person, it sends a signal.

He presented the Universal Commerce Protocol (UCP), Google's response to OpenAI's Agent Commerce Protocol (ACP), which was announced at the end of September 2025.

What many are now seeing: Two tech giants are building similar protocols for shopping in AI assistants.

What most people overlook is that this is not a technical gimmick. It is a battle over standards. A battle that will determine who controls the infrastructure for AI-driven commerce.

And Google has a structural advantage that hardly anyone really understands.

What exactly is UCP?

First, the basics. Without marketing jargon.

UCP is the protocol that enables commerce within Gemini and Google AI Mode.

The user flow is simple:
🟢 You search for a product in Gemini
🟢 You see product tiles directly in the chat
🟢 You can buy directly in the interface
🟢 Checkout is done via an integrated wallet (currently Google Pay)

No more switching to external shops. The entire shopping experience remains within the AI interface.

✅ This isn't a new concept. OpenAI is already doing it with ChatGPT. Amazon is experimenting with it. But the implementation differs fundamentally.

How retailers tie themselves in

Retailers who want to sell via UCP integrate via Google Merchant Center.

That's the first crucial point: Google Merchant Center already exists. It has been powering Google Shopping for years. Millions of SKUs are already onboarded. Product catalogs are available. Structured data exists.

What merchants need to do:
1️⃣ Use existing Merchant Center integration (if available)
2️⃣ Enrich product data for LLMs (not just keyword-optimized)
3️⃣ Implement an "agentic checkout service"
4️⃣ Connect to existing commerce infrastructure

Shopify has already announced that it will support UCP natively. As the main contributor to the protocol. This is not just any partner. This is the backbone for millions of online stores.

The difference to keyword search

This is where it gets interesting.

The product feeds that are currently optimized for Google Shopping are not sufficient. Google has introduced many new product attributes.

Why?

Large language models work differently than keyword searches. They require richer context. They understand connections. They can process nuances.

Simply describing a product as "red dress size M" is not enough. LLMs need material, occasion, style, fit, care instructions, and possible combinations.

In concrete terms, this means that retailers must fundamentally revise their product data, even if they already use Google Shopping.

✅ This is no small undertaking. It is a strategic investment.

UCP is general purpose – theoretically

Google positions UCP as an open, general-purpose protocol.

Not just for Gemini. But for an ecosystem. Many traders should be able to interact with many AI agents. NxN architecture instead of a 1:1 relationship.

UCP even has a built-in discovery mechanism. In theory, merchants can publish their commerce capabilities on their own domain. Agents can then find them in a decentralized manner. Without a central gatekeeper.

But here comes reality:

In practice, most merchants will go through Google Merchant Center for Gemini in 2026. The decentralized discovery mechanism remains theoretical.

Google is building an open protocol. But the implementation runs on Google's infrastructure.

I call this: "Open-source, but I go first."

👉 Do you know of any other examples of this strategy? And how did it develop in the long term?

Checkout mechanism: Conceptually identical to ACP

The actual checkout mechanism at UCP is conceptually identical to OpenAI's ACP.

The flow:
1️⃣ Payment credentials are sent to a processor
2️⃣ Tokenization
3️⃣ Token goes back to the AI platform
4️⃣ Token is forwarded to the checkout gateway
5️⃣ Gateway integrates with merchant systems

Important: The merchant remains the "merchant of record." The AI platform is not part of the payment flow. It only orchestrates.

UCP is agnostic when it comes to payment processors. You are not tied to a specific provider. This is different from ACP, which is heavily focused on Stripe.

UCP also remains neutral when it comes to choosing a commerce tech stack. Shopify, Magento, WooCommerce, custom solutions—anything is possible.

✅ Flexibility is the key argument here. But flexibility also means greater complexity during implementation.

The fundamental differences to OpenAI's ACP

Now it's getting strategically interesting.

Both protocols solve the same problem: they enable checkout directly in AI interfaces.

But the design philosophies couldn't be more different.

Difference 1: General Purpose vs. Platform-Centric

OpenAI's ACP:

🟢 Compactly built
🟢 Optimized around OpenAI and Stripe
🟢 Tightly scoped
🟢 Quick to deploy
🟢 Clearly defined specifications

Google's UCP:
🟢 General purpose
🟢 Ecosystem-oriented
🟢 Designed for NxN connections
🟢 Broader, more ambitious systems
🟢 Future-facing abstractions

OpenAI shipped first, with a narrowly defined specification. Google followed with a more ambitious but also more complex system.

Difference 2: Discovery mechanism

ACP does not have a native discovery mechanism.

If you want to sell as a merchant on ChatGPT, you have to go through a centralized listing process controlled by OpenAI.

UCP theoretically enables decentralized discovery.

Merchants can publish their commerce capabilities on their own domain. Agents can find them without a gatekeeper.

This is fundamentally different in architectural terms. Even if, in practice, both paths will initially run via the respective platform.

Difference 3: Allowances and payment restrictions

ACP has an elegant feature: allowances.

Stripe's Shared Payment Token supports programmable restrictions. As a user, you can define: This token is only valid for 24 hours. Or: Maximum $100 per transaction.

Fine-grained control. Natively built-in. As-a-service.

UCP has no equivalent native concept.

Similar behavior would be possible via AP2 (a former Google Payment Standard). But that is more complex. It requires cryptographic signatures. A more complex tech stack.

Less elegant than ACP's approach.

Difference 4: Risk Signaling

At ACP, risk signals are only sent to the payment processor, i.e., Stripe.

With UCP, risk signals go directly to the trader.

This is closer to current practice. Retailers typically use their own fraud and risk vendors. They want control over risk decisions.

✅ UCP is closer to existing merchant setups here.

Difference 5: Authentication

ACP is currently unauthenticated. Instant checkout without authentication.

UCP allows OAuth-based agent-to-merchant authentication. Merchants can specify that agents must interact in an authenticated manner.

This may not be immediately relevant for OpenAI vs. Gemini (both have session authentication). But as soon as third-party agents come into play, it becomes important.

The strategic advantage that almost no one sees

Now comes the point that, in my view, is most underestimated.

Google Merchant Center.

I cannot emphasize this enough: this is Google's structural advantage.

Product catalogs are the operational bottleneck in agentic commerce. SKUs must be deployed, kept up to date, and enriched.

That's brutally time-consuming. Much more time-consuming than it looks from the outside.

A commenter on the original article wrote: "I worked through a catalog sync nightmare last year and it took way longer than anyone expected."

Google has already solved this problem.

Millions of SKUs have been onboarded. Product feeds exist. Integrations are in place. For Google Shopping. For years.

The feeds are not yet as rich as the new LLM-optimized attributes require. But the foundation is there.

OpenAI does not have this advantage.

You have to build commerce infrastructure from the ground up. Build teams. Build merchant integrations.

This gap means: months. Maybe years.

👉 That is the difference between "protocol announced" and "working system with real GMV."

"Open source, but I go first" as a strategy

Both platforms pursue the same basic strategy:

They publish a protocol. They position it as open. But they promote their own platform first.

OpenAI:

🟢 Released ACP
🟢 Powers ChatGPT first
🟢 Deeply integrated with OpenAI Platform and Stripe

Google:
🟢 Releases UCP
🟢 Powers Gemini and Google AI Mode first
🟢 Deeply integrated into Google Pay and Merchant Center

That's not criticism. That's smart strategy.

But it shows that "open protocol" does not mean neutral governance.

The platform's own implementation will always receive preferred features, better performance, and more direct support.

✅ For merchants, this means: Multi-homing makes sense. Support both protocols. Avoid lock-in.

Direct Offers: The ads layer is coming

Google announced Direct Offers at the same time.

This is not a core UCP feature. But it is a signal.

What Direct Offers enables:

Retailers can send targeted offers to users based on search intent.

Example: User searches for Nike shoes. Adidas can place a direct offer for comparable Adidas shoes at a discount.

Competitor targeting. In the AI interface.

This is the first step toward a complete ads and promotion layer on AI platforms.

Upsells and promotional offers will become a major force in UCP. Not immediately, but in the medium term.

My assessment:

Google's ad business on AI surfaces will expand faster than most people expect.

Promotional placements are becoming the core monetization layer.

And I expect ChatGPT to get an equivalent to Direct Offers in 2026. OpenAI won't stand by and watch Google build up an advertising lead here.

The UCP Roadmap: Where is the journey headed?

Google has published an official roadmap:

1. Loyalty and member benefits Standardized management of loyalty programs and rewards.

2. Native cross-sell and upsell Enhanced Signals for personalized product discovery.

3. Product discovery and post-order management Support for the entire commerce lifecycle. Optimization based on lifetime value rather than single transactions.

4. Cart and basket building Native multi-item checkout. With support for promotions, tax, shipping, and complex fulfillment logic.

These are not minor features. This is a vision for a complete commerce infrastructure within AI interfaces.

✅ But be careful: roadmaps are declarations of intent. They are not guarantees.

What will really happen in 2026

Now it's becoming realistic.

There will be a significant gap in 2026 between the futuristic promises of UCP and what consumers actually use.

In practice, most of it will be:

A "Buy Now" button in Gemini or Google AI Mode.

Decentralized discovery? Theoretically possible, but not used in practice.

Multi-agent ecosystems? They're coming. But not in 2026.

Complex loyalty integration? On the roadmap. Not in production.

What really happens:

🟢 Google is reaching commerce scale faster than OpenAI (despite a smaller Gemini user base).

🟢 Buy Now Flows achieve substantial purchasing volume

🟢 Direct Offers are expanding faster than expected

🟢 OpenAI accelerates ACP opening and offers more attractive merchant terms

🟢 An equivalent to Direct Offers appears in ChatGPT

🟢 UCP and ACP begin to leapfrog each other (Protocol Leapfrogging)

Protocol Leapfrogging: The Standards Dynamic

That's a concept that I find important.

UCP and ACP will not remain static opposites. They will overtake each other.

Features introduced in UCP—authentication, discovery—will quickly appear in some form in ACP.

And vice versa.

This is not a "choose once and stick with it" scenario. This is continuous evolution.

Why?

Because both platforms are under strategic pressure.

OpenAI needs to become more attractive to merchants. Otherwise, they will lose out to Google's Merchant Center advantage.

Google needs to iterate faster. Otherwise, they will lose out to OpenAI's ChatGPT user base.

Competitive standards battles accelerate innovation.

✅ The best features are copied. The bad features die out.

What Meta and Anthropic will do

The battle over standards is not bilateral.

Meta and Anthropic will have to respond in 2026.

The options:
1️⃣ Adopt UCP
2️⃣ Adopt ACP
3️⃣ Own protocol

My assessment: At least one of these platforms will build its own protocol.

Why?

Because commerce infrastructure means strategic control. Whoever controls the protocol controls merchant relationships. The data. Monetization.

👉 What do you think: adopt or create your own protocol?

What this means for retailers

Now for the practical implications.

If you are a retailer who takes agentic commerce seriously, you have three decisions to make:

Decision 1: Timing

When are you coming on board?

Early movers are likely to get better economics. Better placements. More direct support.

But early movers also bear the risk. Protocols change. Standards consolidate.

My assessment: If you already use Google Merchant Center, the risk is manageable. UCP builds on existing infrastructure.

When you start at OpenAI, the risk is higher. But the ChatGPT user base is larger.

Decision 2: Scope

UCP, ACP, or both?

For UCP, speak:
🟢 You already use Google Merchant Center
🟢 You rely on the Google ecosystem
🟢 You are planning for a multi-agent future

ACP is a good choice if:
🟢 You want to sell quickly in ChatGPT
🟢 Stripe integration is okay with you
🟢 You prefer clearly defined specs

Arguments in favor of both:
🟢 You want to avoid lock-in
🟢 You have resources for double integration
🟢 You want maximum reach

My recommendation: Start with one. Learn. Then expand.

Decision 3: Product data strategy

This is often underestimated.

Product data enrichment is not optional. It is mandatory.

LLMs require richer context than keyword searches. The existing feeds optimized for Google Shopping are not sufficient.

What you need to do: 1️⃣ Revise product descriptions for LLMs 2️⃣ Add new attributes (material, occasion, style, fit, etc.) 3️⃣ Expand structured data 4️⃣ Ensure continuous maintenance

It's an investment. But it pays off. For both protocols.

My personal assessment

After 10+ hours of research and analysis:

Google has a structural advantage.

Merchant Center is the game changer. The operational bottleneck—product catalog onboarding—has already been resolved.

This advantage is difficult to catch up with. Months, maybe years.

But OpenAI has momentum.

ChatGPT has the larger user base. Earlier on the market with ACP. Clearer, simpler specs.

Both will coexist.

This is not a "winner takes all" situation. It is an ecosystem with multiple players.

UCP and ACP will coexist. Features will converge. Standards will become more similar.

The real question:

How are open vs. closed standards developing in the AI commerce landscape?

Google is positioning itself as more open. With decentralized discovery. With general-purpose design.

But in practice, everything runs through Google Merchant Center.

OpenAI is more honest. They are platform-centric. But that's exactly why they can be faster.

What I will observe:

🟢 How quickly will Google achieve real GMV with UCP?
🟢 How attractive are OpenAI's terms for merchants?
🟢 When will Meta and Anthropic come up with their own solutions?
🟢 How quickly will the protocols converge?
🟢 What role does decentralized discovery really play?

The open questions

Some things remain unclear:

1. Merchant economics How do the economics differ in concrete terms? What are the fees? What are the margins? The article does not provide any specific figures.

2. Decentralized discovery Is this being used in practice? Or does Merchant Center remain the de facto gatekeeper?

3. OpenAI's catch-up strategy What specific incentives is OpenAI using to catch up with merchant integrations?

4. Meta and Anthropic What exactly do they do? When?

5. ONDC/beckn Protocol How does UCP relate to ONDC and beckn Protocol? (A question that was asked in the comments but has not been answered.)

Conclusion: The battle over standards has begun

Agentic commerce is no longer a theoretical concept.

It's a measurable race. With billions of dollars in purchasing volume as the finish line.

Google and OpenAI have placed different bets:

Google: Infrastructure advantage + ecosystem design + existing merchant relationships

OpenAI: Speed + simpler specs + larger user base

Both approaches are coherent. Both are justified.

What 2026 will show: Does Google's structural advantage outweigh OpenAI's momentum?

For retailers, this means: Decide now. When to onboard? UCP, ACP, or both? How to enrich product data?

These decisions shape distribution, economics, and visibility on AI-native shopping surfaces.

This is not hype. This is strategic reality.