How to prepare your store for agentic checkout.

The slow way: read the specs, audit yourself

The manual path is real but demanding. Read the protocol material — OpenAI and Stripe's ACP documentation, Google's UCP announcements, the AP2 payments work — and map which surfaces each one actually buys you. Then audit your own store the way an agent would: is the price machine-readable on first fetch, or rendered by JavaScript? Can a crawler reach the returns policy by URL? Is the Product JSON-LD complete enough to draft an accurate cart line — variant, price, availability, shipping class?

The audit half is a long afternoon per template plus per-SKU spot checks. The protocol half is worse, because the material is vendor-written and changes quarterly: tier eligibility, partner lists, and rollout schedules shift, and a decision memo written in March needs rewriting by June. Most teams either over-invest (a custom ACP backend nobody asked for) or freeze. The middle path — groundwork now, protocol decision on a quarterly review — rarely emerges from vendor docs.

The eCommerce Insights way

1. Run the agentic-readiness assessment. The Agentic Readiness Grader scores the groundwork every protocol presumes: Product JSON-LD completeness, machine-readable price and availability, discoverable policies, crawler admittance. Free, per URL, about 30 seconds.
2. Fix the groundwork first. Whatever happens to the acronyms, draft-cart agents — ChatGPT Shopping, Perplexity Buy with Pro — parse PDPs today. Gaps ship as reviewable diffs through the standard PDP optimization workflow; the same fixes raise both the agent-readability and citation scores.
3. Map the protocol options. The assessment report sorts them honestly: Shopify brands get the most reach for the least work through Shopify's agentic storefront channel, which wraps the OpenAI relationship; ACP direct integration is a six-to-twelve-week engineering project justified only by a strong ChatGPT-specific case; UCP rides Google's surfaces; AP2 is the payments layer underneath. Every maturity claim hedged — these are pilots as of mid-2026.
4. Decide the SKU opt-in policy. Agent checkout is a per-SKU decision, not a store-wide toggle. Exclude subscriptions, age-restricted, services, and fulfillment-fragile items first; opt in clean high-velocity SKUs and widen as order flow proves out.
5. Monitor and revisit quarterly. The agent-readability score per SKU is tracked on every refresh, and the protocol landscape gets re-read each quarter. This quarter's right answer expires; the groundwork doesn't.

The full architecture — what each protocol actually specifies, who backs it, what merchants report — is in the ACP guide and the UCP guide, with the strategic frame in agentic commerce.

What "good" looks like

The realistic 2026 posture: groundwork complete, one platform-mediated channel enabled where eligible, no custom protocol engineering without a commercial case. Brands holding that posture capture the draft-cart volume that exists now and are first in line if agent-completed checkout scales.

Related jobs