Skip to content
UCP
Menu

Thesis

The commerce layer AI agents are going to read

A new infrastructure layer is forming between online commerce and AI systems. Here is the editorial thesis that organizes everything we publish.

Updated : April 2026 · Primary query : future of commerce ai agents

A new infrastructure layer is forming between online commerce and AI systems. Its purpose is to make catalogs, offers, stock, policies and intents readable, reasonable-about and actionable by agents as well as humans. We call this layer the Universal Commerce Protocol — a conceptual frame, not a single spec. The merchants, platforms and tools that adopt UCP-shaped patterns early will be the ones retrievable, recommendable and transactable by the next generation of shopping interfaces.

Claim 1 — Humans are no longer the only clients of commerce data

Status: established.

For two decades, product pages, feeds and APIs were optimized for two consumers: human shoppers and search engines (essentially indexing surfaces for humans). AI agents are a third, structurally different consumer — they read at machine speed, compare across catalogs, reason about intent, and act on behalf of the human. Optimizing only for humans and SERPs is a position that is already obsolete.

Evidence: ChatGPT Shopping, Perplexity Shop, Amazon Rufus, Google AI Overviews commerce surfaces are all in production. Each reads product data on behalf of a user.

Claim 2 — The commerce stack is missing a protocol layer

Status: emerging.

There is no shared contract for how an agent asks "does this merchant sell a rain jacket under €120 that ships to Berlin by Friday, and what are the return policy and warranty terms?" — and gets a deterministic, structured, transactable answer. Today the agent scrapes HTML, queries a patchwork of feeds, or relies on a walled-garden marketplace. This is fragile, expensive, non-interoperable. A protocol layer — a set of shared semantics — will emerge by necessity, the way OAuth, OpenAPI and schema.org emerged in their respective moments.

Claim 3 — The protocol will not be a single spec

Status: probable.

"Universal Commerce Protocol" is a conceptual convergence, not a W3C document. Expect composite emergence: payment rails (Stripe ACP, Visa IC, Mastercard Agent Pay), agent tooling (MCP, Apps SDK, A2A), catalog semantics (schema.org evolution, GS1, Google Merchant Center), identity (agent-as-user credentials), and trust layers (provenance, authenticity). The site's role is to map, name and synthesize this composition.

Claim 4 — Merchant readiness is the decisive variable

Status: emerging, becoming established.

Much of the stack will arrive pre-integrated by platforms (Shopify, WooCommerce, BigCommerce, Adobe Commerce, Salesforce Commerce Cloud). But catalog quality, attribute richness, policy clarity, and structured-data hygiene remain merchant responsibilities. The gap between "platform-ready" and "agent-retrievable" is mostly owned by the operator. This is the operational wedge the site occupies — see the readiness checklist.

Claim 5 — Post-SERP discovery changes the economics of content and merchandising

Status: emerging.

When agents mediate discovery, traditional SEO becomes a precondition, not a finish line. The next frontier is semantic addressability: being the answer, not the link. Brands that fail to render themselves as machine-readable propositions risk invisibility in agent-first surfaces — Perplexity Shop, Rufus, ChatGPT Shopping, Comet, and successors yet unnamed.

How the thesis branches across the site

Thesis claimPrimary page
Humans aren't the only readersWhy machine-readable
Commerce needs a protocol layerWhat is UCP
No single spec; composite emergenceStandards map
Agents as a new commerce clientAI agents & commerce
Interoperable catalogs and semanticsInteroperability, Catalogs for AI
Merchant operational wedgeReadiness checklist, Audit
Post-SERP economicsFuture roadmap

What we deliberately avoid

  • Claiming to be a standard-setter.
  • Predicting winners.
  • Timelines in calendar years. We use horizons instead (near, medium, long).
  • Speculation presented as established fact — every claim is labelled.

Refresh cadence

  • Thesis reviewed every 6 months.
  • Glossary continuously.
  • Standards comparison quarterly.
  • Case studies as signals emerge from the field.