ReadinessUpdated 2026-08-17

Is your store ready for agentic commerce?

Agentic commerce is the shift from a person browsing your storefront to an AI agent doing the discovery, comparison and increasingly the purchase on their behalf. Agentic commerce readiness is whether your store survives that shift — whether an agent can reach your catalogue, parse your product data, and justify choosing you over a competitor it can parse better.

Definition

Agentic commerce readiness is a merchant's operational readiness for transactions initiated by AI agents: machine-readable product and policy data, crawl access for agent user-agents, and product information complete enough for an agent to defend the recommendation.

42%
of merchants already testing agentic commerce (Checkout.com, 2026)
36%
of stores we scanned expose machine-readable product data
22%
publish a sitemap containing no product URLs at all

What actually changes for a merchant

Agentic commerce is not a new sales channel you sign up for. It is an existing channel — search, comparison, discovery — with the human removed from the middle of it. That removes three things merchants have relied on for twenty years.

The page stops mattering
Layout, photography, badges and social proof are inputs to a human decision. An agent reads your markup. If the price lives only in rendered HTML, the agent sees a page with no price.
The objection gets answered without you
A shopper who cannot find your return policy emails you. An agent that cannot find your return policy recommends the store whose return policy it could read.
The comparison happens off-site
You are not being compared on your product page any more. You are being compared inside a model's context window, against competitors you never see, on fields you may not have filled in.

Agentic commerce, agent commerce, AI commerce

The vocabulary is still settling. "Agentic commerce" has become the category name — it is what McKinsey, Stripe, Mastercard, Salesforce and the Agentic Commerce Protocol use. "AI commerce" is the broader umbrella that includes merchandising, support and forecasting. "AI shopping agent" names the actor rather than the category.

Prefero measures readiness for the actor. Whatever the category ends up being called, the merchant-side work is the same: make the catalogue machine-readable, keep the crawl paths open, and state policy in a form a machine can quote.

TermWhat it namesMerchant question it raises
Agentic commerceThe category: transactions where an AI agent acts for the buyerAm I present in this channel at all?
AI shopping agentThe actor: ChatGPT shopping, Perplexity, Gemini, browser agentsCan this specific agent read my catalogue?
Agentic Commerce ProtocolOne emerging standard for agent-initiated checkoutWill my checkout accept an agent-initiated order?
AI commerce readinessThe measurable merchant-side stateWhere exactly am I losing, and what do I fix first?
How the terms divide up

The readiness checklist

In order of how often it is the actual blocker in the stores we scan. Fix top-down; there is no value in tuning aggregate ratings on a catalogue an agent cannot enumerate.

1. Product URLs are discoverable
A sitemap that resolves and contains product URLs. Twenty-two percent of the stores we scanned published a sitemap with none.
2. AI user-agents are not blocked
Retrieval crawlers — OAI-SearchBot, ChatGPT-User, PerplexityBot, Claude-User, Googlebot — are the ones that put you in front of a buyer. Blocking them is usually accidental, inherited from a bot-mitigation rule.
3. Product JSON-LD is server-rendered
Client-injected schema is invisible to most agent fetches. If your theme writes JSON-LD from JavaScript, assume it does not exist.
4. Availability and price are current
A stale availability field is worse than a missing one: an agent that recommends an out-of-stock product learns not to trust the source.
5. Returns and shipping are machine-readable
hasMerchantReturnPolicy and shippingDetails on the offer, not a link to a policy page written in prose.
6. Identity is unambiguous
brand, gtin or mpn, and a stable sku. Without an identifier an agent cannot confirm your listing is the same product it found elsewhere.

An honest read on timing

Checkout.com's 2026 agentic commerce report puts 42% of merchants at the testing stage and roughly nine in ten saying they are preparing. That is real interest, not adoption. Meanwhile early agentic checkout flows have been revised after weak conversion results, and the standards are not settled.

The reasonable position is neither "agents will replace ecommerce next year" nor "wait and see". It is that the channel is forming, the entry cost right now is a weekend of structured-data work, and the same work pays for itself in classic search regardless of how fast agents arrive.

Frequently asked questions

What is agentic commerce?
Agentic commerce is commerce where an AI agent performs the discovery, comparison and increasingly the purchase on a buyer's behalf, rather than the buyer browsing storefronts directly. The merchant is evaluated through machine-readable product and policy data rather than through the rendered page.
What does agentic commerce readiness mean for a merchant?
It means an AI agent can reach your product pages, parse complete product data from them, resolve shipping and return terms without human interpretation, and defend recommending you over an alternative. Prefero scores those four things and returns the specific gaps.
Do I need to implement the Agentic Commerce Protocol?
Not to be discoverable. Protocols like ACP standardise agent-initiated checkout — the last step. Discovery, comparison and recommendation happen through ordinary crawl access and structured data, which is where almost every store we scan is losing today.
Is agentic commerce actually driving revenue yet?
Volumes are still small and early checkout experiments have been reworked after weak conversion. The case for acting now is cost, not volume: the readiness work is inexpensive, it overlaps with existing SEO and feed hygiene, and the alternative is discovering you are unreadable after the channel matters.
Which agents does Prefero test against?
The crawl-access check covers the published retrieval and training user-agents from OpenAI, Google, Perplexity, Anthropic, Apple, Amazon and Meta. The buyer simulation runs model-driven shopper personas against your product data and a competitor's.

See where you stand on agentic commerce

One URL, one score, the ranked list of what is blocking you. Free and no account required.

Your store vs. competitor

Unlocks AI preference

Free scan · ~30 seconds · No account

Related

From the blog