Agentic Commerce

In agentic commerce, an AI agent acts on behalf of a shopper across part or all of the buying journey: understanding what the person wants, searching across stores, comparing options against stated preferences and constraints, and in some cases placing the order. It spans a spectrum from assistive, where the agent researches and recommends and the person clicks buy, to autonomous, where the person approves an intent and the agent completes the purchase within set limits. The common thread is that a machine, not a human browsing a storefront, does much of the discovery and evaluation.

A set of open standards is forming to make this work, and they are designed to compose rather than to crown a single winner. The Agentic Commerce Protocol (ACP), co-developed by OpenAI and Stripe, defines how an agent completes a checkout with a merchant. The Agent Payments Protocol (AP2), led by Google with 60+ launch partners, adds a payments and authorization layer that cryptographically ties a buyer's intent to the merchant's charge. The Model Context Protocol (MCP), from Anthropic, is the underlying tool and context layer that lets an agent read data and call the APIs those layers expose. A single purchase can touch several of them at once.

The current state is real but early and uneven. Discovery and recommendation, products surfacing inside AI answers and being compared against a shopper's needs, is the mature part today. One-click checkout inside the assistant is still nascent: OpenAI launched Instant Checkout, then narrowed it around early 2026 after only a small set of Shopify merchants went live and accurate product data proved hard to get right at scale. As of mid-2026 products appear inside engines like ChatGPT, but most purchases still complete on the merchant's own storefront. The durable takeaway is that the shift is strategic while the winning checkout standard is unsettled, so the safe merchant move is agent readiness rather than a bet on one flow.

For a store, the stakes are simple: an agent can only buy what it can first find and trust. Whether your products show up in an AI shopping answer, and whether an agent is confident enough to recommend or transact them, is decided upstream by the quality of your product data and your visibility across AI engines. Accurate, structured, machine-readable catalog information, correct price, availability, variants, and identifiers, plus a consistent presence in AI results, is what earns a product a place on the shortlist an agent works from. Thin or mismatched data quietly leaves it out.

This is the layer a Shopify merchant actually controls. Whichever checkout protocol wins, the prerequisite is the same: a clean, transactable catalog and strong AI visibility. A tool like Naridon focuses on that upstream readiness, tracking whether AI engines cite your products and applying structured-data and content fixes, rather than on payments or checkout, which stay with the protocols and your existing payment provider.

Illustrative scenario: a shopper tells an AI assistant "find me a waterproof running jacket under $150 in medium, in stock, ships to Canada." The assistant queries several stores, filters by the stated constraints, and returns a short list. A store whose product data exactly matches those facts, price, size, availability, and shipping, is easy for the agent to include and recommend; a store with a vague title and stale stock is easy to skip, no matter how good the product is.