Agentic Commerce Optimization (ACO) & UCP Protocol

Definition and context

Agentic Commerce Optimization (ACO) is the practice of structuring product data, signals, and technical infrastructure so products are selected by autonomous AI agents. As the web shifts from human-led search to AI-mediated selection, ACO prioritises data integrity and machine-readability over visual persuasion.

The shift is driven by the Universal Commerce Protocol (UCP), a framework that extends commerce beyond checkout into discovery, loyalty, and post-purchase support. Where the narrower Agentic Commerce Protocol (ACP) covers the transactional layer (checkout → fulfilment → payment), UCP spans the whole lifecycle, letting an agent integrate once and interact with many platforms. Early adopters such as Wayfair and Etsy have begun integrating UCP to enable direct purchasing within Google’s AI Mode.

The six core capabilities of UCP

UCP defines six layered capabilities agents use to navigate the commerce lifecycle. Full agent compatibility means addressing each.

Product discovery — how agents find, parse, and surface inventory during research.
Cart management — multi-item baskets, dynamic pricing, and complex basket rules.
Identity linking — OAuth 2.0 authorization for personalization and loyalty integration.
Checkout — session creation, tax calculation, and payment processing.
Order management — webhook-based lifecycle updates and logistical tracking.
Vertical capabilities — extensible modules for specialised cases (e.g., travel windows, subscription schedules).

Technical requirements: schema and data

UCP uses its own versioned JSON schema for transactions, but Schema.org remains the critical glue for discovery — agents use standard schema to decide who to transact with before UCP handles how.

Mandatory product schema fields

Core identity: name, description, SKU, GTIN, brand, image.
Offers: price, priceCurrency, availability, url, seller.
Trust signals: aggregateRating and review for third-party validation.
Logistics: shippingDetails with precise delivery estimates.
Variants: every variant (size, colour, material) explicitly defined.

Organization and support schema

Merchant of record: Organization (fallback Person) to verify who is responsible for the transaction.
FAQPage: required to support agent decision logic on policies and product details.

Merchant Center configuration

UCP uses the Google Merchant Center feed as its primary discovery layer, and requires attributes beyond the standard feed.

Critical feed attributes

native_commerce — assert this to declare a product eligible for agentic checkout.
Product identifiers — strict correlation between the feed ID and the product ID used in the Checkout API.
Consumer notices — any product warnings use the consumer_notice attribute.

Policy and support data

Return policies — complete data on costs, windows, and policy links, so the agent can validate the merchant of record.
Customer support — structured support data lets agents handle basic (L1) queries autonomously.

Implementation note: Google recommends adding UCP-specific attributes via a supplemental feed to avoid formatting errors in the primary feed.

Conversational commerce attributes

To reduce hallucination during discovery, UCP adds “conversational commerce attributes” — data points that let agents parse product relationships standard keywords miss.

Compatibility — what the item works with (e.g., “compatible with iPhone 15 Pro”).
Substitution — acceptable alternatives for out-of-stock scenarios, to prevent abandonment.
Related products — structured data for cross-selling.

Agents also need high-resolution descriptors: supplement generic attributes (Colour: “Purple”) with specific, parseable detail (Colour: “Wolf” or “Dark Slate”) to satisfy long-tail queries.

Optimizing for ACO increases visibility in fan-out queries, where one prompt (e.g., “plan a camping trip”) decomposes into sub-intents (tent, sleeping bag, cooking gear).

Visual fan-out — agents analyse a single image and fan out to find all constituent products.
Attribute priority — agents favour products with conversational attributes (compatibility, substitution) when assembling bundles from one prompt.

Future roadmap

UCP is expected to expand beyond retail into service verticals:

Complex baskets — native bundling, promotions, and multi-item carts.
Loyalty integration — standardised linking for points and member pricing.
Post-purchase agents — managing returns and tracking without human intervention.
New verticals — travel, digital goods, and food/restaurant services.

Agentic SEO — the broader agentic-readiness discipline
GEO: Getting Cited in AI Answers