Google’s Universal Commerce Protocol (UCP)

By /
UCP

UCP is a big deal structurally, but still early and slightly overhyped tactically. What you’re seeing (including that “agentic sizing platform”) is exactly the first wave of people trying to monetize the shift before the rails are fully built.

Let me break it down in a way that maps to how you think about domains / infrastructure bets.

🧠 What Google’s UCP actually is (under the hood)

At its core, Google’s Universal Commerce Protocol (UCP) is:

  • An open standard for AI-driven commerce
  • A shared language for agents to transact, not just recommend
  • Designed to handle the full flow:
    👉 discovery → comparison → checkout → payment → post-purchase

And importantly:

  • It plugs into things like MCP (Model Context Protocol), APIs, and agent-to-agent systems
  • It’s backed by a who’s who of commerce: Shopify, Walmart, Target, Stripe, Visa, etc.

👉 The real unlock:
AI doesn’t just suggest products anymore — it can complete the purchase on your behalf.

⚔️ The real framing (this is the important part)

This is not a “feature.”
This is a standards war.

Think:

  • MCP → standard for AI ↔ tools/data
  • UCP → standard for AI ↔ commerce

Together, they form the rails for agent autonomy.

Or more bluntly:

UCP is trying to become the “HTTP of buying things via AI.”

🧩 Why this matters (and why you’re right to pay attention)

1. It compresses the funnel to zero

No more:

  • search → click → browse → cart → checkout

Instead:

  • “buy me the best X under $100” → done

That’s existential for:

  • SEO
  • affiliate
  • marketplaces
  • even brand websites

2. It shifts power to whoever controls the agent layer

Right now:

  • Google is trying to own it via Gemini + Search
  • OpenAI is pushing similar ideas (ACP)
  • Amazon will absolutely play here

👉 The interface becomes the gatekeeper again.

3. It turns product data into the new SEO

The dirty secret in UCP:

  • Whoever has the cleanest, richest product feed wins
  • Not whoever has the best website

That’s why people are already building tooling around:

  • feed optimization
  • pricing signals
  • margin-aware ranking

👀 Your “agentic sizing platform” example (this is interesting)

That’s actually a perfect early wedge.

If Gap is involved, that tells you something:

👉 Apparel is one of the hardest categories for agents
(because sizing, fit, returns = chaos)

So what that company is doing is:

  • Owning a decision layer inside the agent flow
  • Not competing with UCP — plugging into it

Think of it like:

“We don’t sell the shirt — we decide which shirt fits.”

That’s valuable because:

  • Agents need confidence signals to act autonomously
  • Sizing is a high-friction blocker

So yeah — that’s a legit play.

⚠️ The skepticism (where I’d push back)

1. “Open standard” doesn’t mean neutral

Google says open…

…but:

  • they control distribution (Search, Gemini)
  • they influence ranking logic

So:

UCP could become “open in theory, Google-controlled in practice”

2. Adoption friction is real

Retailers have to:

  • expose inventory + pricing cleanly
  • trust AI to transact
  • give up some UX control

That’s a big psychological and operational leap.

3. Consumer behavior isn’t fully there yet

People still:

  • browse
  • compare emotionally
  • distrust full automation for purchases

So near-term:
👉 hybrid (assistive AI) > fully autonomous buying

💰 From an opportunity perspective

This is where it gets fun.

You’re early if you’re thinking in terms of:

Categories that will explode:

  • Agent optimization
    • (like SEO → but for AI decision engines)
  • Trust layers
    • reviews, validation, authenticity signals
  • Decision engines
    • sizing, compatibility, personalization
  • Feed infrastructure
    • data normalization, enrichment
  • UCP = very real, foundational shift
  • But:
    • Not immediate
    • Not winner-take-all yet
    • Still forming standards + power structure

If MCP was:

“AI can use tools”

Then UCP is:

“AI can spend money”

That’s a different level entirely.

⚔️ PART 1 — UCP vs ACP vs Amazon (the real war)

🟦 Google UCP (Universal Commerce Protocol)

Positioning:

“We standardize how AI buys things across the internet.”

Strengths:

  • Distribution via Google (Search, Shopping, Android)
  • Existing merchant ecosystem (millions already in feeds)
  • Payments + identity hooks (Google Pay, accounts)

Strategy:

  • Make commerce agent-native
  • Turn product feeds into machine-readable APIs
  • Own the decision layer via Gemini

Weakness:

  • Historically bad at owning checkout
  • Merchants don’t fully trust them

🟩 OpenAI ACP (Agent Commerce Protocol – emerging concept)

(This is less formally defined than UCP, but very real directionally)

Positioning:

“The agent is the interface. Commerce happens through it.”

Strengths:

  • Interface dominance via OpenAI / ChatGPT
  • Deep integration with MCP (tools, APIs, actions)
  • User trust in conversation-driven decisions

Strategy:

  • Don’t standardize the web — abstract it
  • Agents call APIs, not browse sites
  • Commerce becomes intent → action

Weakness:

  • No native merchant network (yet)
  • Relies on integrations (Stripe, Shopify, etc.)

🟧 Amazon (the sleeping giant here)

Positioning:

“Why leave our ecosystem at all?”

Strengths:

  • Amazon owns:
    • inventory
    • logistics
    • checkout
    • trust
  • Prime = built-in loyalty + payment layer

Strategy (likely):

  • Keep agents inside Amazon
  • Build their own agent interface (Alexa+, Rufus, etc.)
  • Offer best price + fastest delivery → win by default

Weakness:

  • Closed ecosystem
  • Doesn’t benefit from “open standards” like UCP

🧠 The clean mental model

  • UCP (Google) → “Let’s standardize the open web”
  • ACP (OpenAI) → “Let’s abstract the web entirely”
  • Amazon → “Let’s ignore the web and win anyway”

🔥 The real battleground

Not APIs. Not protocols.

👉 Who controls the agent that makes the decision.

Because:

Whoever controls the agent controls the transaction.

Here is an example

Agentic AI that a retailer can plug into their website and do a better job of making sure a customer chooses the correct size of, for example, a polo shirt.

👕 The problem it solves

Clothing has:

  • insanely high return rates (20–40%+)
  • inconsistent sizing across brands
  • subjective fit (“slim”, “relaxed”, etc.)

For AI agents, this is a nightmare:

An agent can’t confidently buy if it thinks it’ll be wrong.

💡 What that platform is really doing

It’s not a “sizing tool.”

It’s a:

confidence engine for autonomous purchasing

It likely:

  • maps body data → brand sizing
  • predicts fit outcomes
  • reduces return probability

🧠 Why that matters in an agent world

Agents need:

  • certainty signals
  • risk reduction
  • decision confidence

Without that:
👉 they hesitate or defer to the user

With that:
👉 they execute

🧱 Where it sits in the stack

Think of the future stack like this:

  1. Agent (ChatGPT / Gemini / Alexa)
  2. Protocol (UCP / ACP)
  3. Decision modules (THIS is where sizing lives)
  4. Merchant / inventory
  5. Payment + fulfillment

That sizing company is:

👉 Layer 3 — and that’s a goldmine layer

⚠️ The risk in that model

Here’s the catch:

  • If UCP or Amazon builds native sizing intelligence…
    👉 they get commoditized

So their moat must be:

  • proprietary data
  • cross-brand intelligence
  • better predictions than anyone else

Examples that will exist:

  • sizing (clothing)
  • compatibility (electronics, parts)
  • authenticity (luxury goods)
  • pricing intelligence
  • taste / personalization engines

My blunt take

  • UCP = infrastructure play
  • ACP = interface dominance play
  • Amazon = vertical monopoly play

And the “agentic sizing platform”?

👉 That’s a picks-and-shovels play inside the gold rush

🔮 The one sentence thesis

We are moving from “websites competing for clicks” to “algorithms competing to be chosen by agents.”

Scroll to Top