Google UCP Merchant Readiness Guide: How to Prepare Your Catalog

Universal Commerce Protocol (UCP) can connect Google shopping surfaces to merchant commerce capabilities. It cannot make an unreliable catalog safe for an agent to use. Before investing in endpoints and checkout flows, merchants need to know whether their stack can identify the exact product, expose the current offer, apply the right policy, and explain which source supports each fact.

This guide is the implementation companion to What Is Google’s Universal Commerce Protocol?. It focuses on the operational question: what must be ready for a merchant to participate reliably?

What changed in 2026

UCP moved from protocol announcement toward merchant onboarding and testing for Google’s AI shopping surfaces. The implementation path now spans Merchant Center preparation, a merchant UCP profile, Google Pay, native checkout endpoints, and optional capabilities such as cart transfer, account linking, and order synchronization. Google positions Merchant Center as the central place to prepare product data for its commerce surfaces.

That creates three separate workstreams which should not be collapsed into one:

• Distribution: making products and offers legible in Merchant Center, AI Mode, and Gemini.
• Integration: publishing capabilities and connecting carts, checkout, identity, and orders.
• Data readiness: ensuring every response rests on a resolved, complete, current product record.

Sandbox and integration testing can prove that requests and responses conform. They cannot prove that a GTIN belongs to the right variant or that a supplier dimension is authoritative.

The UCP architecture

The trusted layer should separate three kinds of information:

Do not copy all three into a new UCP-only database. Keep each authoritative system accountable, then expose a governed projection through the protocol.

What UCP does not fix

UCP can transport product, offer, inventory, shipping, policy, checkout, and identity information. It cannot decide which of three duplicate records is canonical, repair a malformed MPN, infer an ambiguous pack quantity, reconcile incompatible units, or establish which supplier value deserves trust.

UCP readiness checklist

1. Identity

• Assign a stable canonical product identity distinct from every supplier or channel alias.
• Model the product family, variant, sellable SKU, and merchant offer as separate entities.
• Validate GTIN checksums, normalize MPNs and brand names, and retain source identifiers.
• Represent bundles, replacement parts, substitutes, accessories, and supersessions explicitly.

2. Attributes

• Normalize values and units into typed fields instead of leaving specifications in prose.
• Measure completeness against category-specific decision attributes, not a universal field count.
• Expose the facts shoppers use conversationally: compatibility, dimensions, use environment, performance, and constraints.
• Keep display copy separate from the normalized values used for filtering and validation.

3. Commercial state

Validate base price, sale price and effective dates, availability, location-level inventory, lead time, shipping constraints, return rules, taxes, and the market or customer context in which each value applies. Treat freshness as data: include an observed or updated timestamp and define the maximum acceptable age per field.

4. Account context

Keep loyalty status, customer eligibility, personalized offers, contract pricing, and regional constraints scoped to the authenticated account. Test the anonymous state separately so private entitlements never leak into a public response.

5. Trust

For every consequential value, preserve its source, source authority, validation result, confidence, and freshness timestamp. Confidence should determine workflow: a low-confidence compatibility claim should route to review or a safe refusal, not become persuasive copy.

6. Operations

Test cart creation and mutation, totals, checkout eligibility, payment handoff, order updates, returns, idempotency, partial failure, timeouts, and error messages. Every operation should preserve the exact product, variant, offer, account, and market selected during discovery.

Test the complete journey, not only the endpoint

Use a constrained scenario such as:

Find the exact replacement filter compatible with model X and buy two.

The stack must perform an evidence chain:

1. 1
   Resolve the installed model
   Map “model X,” including aliases, to a canonical equipment identity.
2. 2
   Verify compatibility
   Find an explicit, sourced relationship to the replacement filter rather than inferring from similar names.
3. 3
   Select the sellable SKU
   Resolve variant, pack quantity, market, and merchant offer without confusing product and listing IDs.
4. 4
   Establish current state
   Confirm price, quantity-two availability, delivery constraints, and the timestamp of those facts.
5. 5
   Execute or fail safely
   Create a valid cart, or return an actionable refusal when a required fact is missing.

Repeat that pattern for a loyalty offer, regional restriction, return-policy exception, substitute, out-of-stock equivalent, and customer-specific price. A successful HTTP response is not a successful test if it contains the wrong item.

Who owns what

Ownership should be explicit because a protocol error and a data error need different remediation. The integration team fixes an invalid operation; the catalog team resolves an uncertain identity; the commerce platform corrects a pricing or fulfillment rule.

The Claro point of view

UCP readiness is not an API project. It is an API project sitting on top of a catalog-readiness project. Claro resolves duplicate product identities, normalizes category attributes, validates conflicting values, preserves provenance, and keeps approved records synchronized with the PIM, ERP, and supplier sources that already run the business.

That work creates a reusable product layer. Merchant Center and UCP can consume it today; other agents and commerce channels can use the same trusted facts without another catalog rebuild.

Make one catalog ready for every agent channel

Claro can audit a representative catalog sample for product identity, variant structure, attribute completeness, freshness, and conflicting values. The output is a prioritized remediation backlog tied to realistic discovery and transaction tests—not a generic score.

See how Claro prepares agent-ready product data.

Related resources

Further reading

• Google Developers: Universal Commerce Protocol
• Universal Commerce Protocol project

FAQ