Agentic commerce, without the hype
Agentic commerce is commerce where an AI agent handles discovery, comparison and increasingly the transaction on a buyer's behalf. The category has real infrastructure behind it — protocol work from OpenAI and Stripe, payment rails from Mastercard and Visa, merchant pilots at scale — and a genuinely uneven track record so far. This page separates the two.
Agentic commerce is a transaction model in which an autonomous AI agent performs product discovery, comparison and purchase steps on behalf of a human buyer, interacting with merchants primarily through machine-readable data and programmatic interfaces rather than a rendered storefront.
The four layers of the stack
"Agentic commerce" is used to describe four different things, which is why the discussion is often confused. They mature at different speeds and demand different things from a merchant.
| Layer | What it covers | Maturity | Merchant work required |
|---|---|---|---|
| Discovery | Agent finds and enumerates candidate products | Working today | Crawl access, sitemaps, indexation |
| Comparison | Agent parses and ranks candidates | Working today | Complete Product structured data |
| Decision support | Agent resolves risk and answers objections | Working today | Machine-readable returns, shipping, warranty |
| Transaction | Agent initiates or completes checkout | Early and contested | Protocol integration, agent-tolerant checkout |
What is real today
The first three layers are not forecasts. If someone asks an assistant for a product recommendation right now, discovery, comparison and risk resolution all happen, and a merchant with unreadable product data loses that comparison silently.
Checkout.com's 2026 report puts 42% of merchants at the testing stage for agentic commerce and around nine in ten preparing for it. That is a signal about merchant attention, not transaction volume — and attention is what the first three layers require.
What is not real yet
The transaction layer is genuinely unsettled. Competing standards exist, payment rails are being defined in parallel by card networks and platforms, and early agentic checkout flows have been revised after weak conversion results.
A merchant who rebuilds checkout today for a standard that has not consolidated is taking a real risk. A merchant who publishes complete product structured data is not — that work is the same work that improves classic search, shopping feeds and comparison-site accuracy.
- Low-risk, do now
- Crawl access for retrieval agents, product sitemaps, server-rendered Product JSON-LD, machine-readable policy fields.
- Watch, do not build yet
- Agent-initiated checkout protocols, agent-specific payment tokens, marketplace listings inside assistants.
- Measure continuously
- Whether your product data is complete and current — this decays as catalogues change, which is why readiness is a monitoring problem rather than a project.
The defensible position
The honest pitch for agentic commerce is not that agents will replace storefronts. It is that a channel is forming, the cost of being legible in it is a few days of structured-data work, and the cost of being illegible is invisible — you never see the comparisons you lost.
That asymmetry, not the growth curve, is the argument.
Frequently asked questions
- What is agentic commerce?
- Agentic commerce is a transaction model in which an autonomous AI agent performs product discovery, comparison and purchase steps on behalf of a human buyer, interacting with merchants primarily through machine-readable data rather than a rendered storefront.
- How is agentic commerce different from ecommerce?
- In ecommerce the buyer evaluates your storefront. In agentic commerce an agent evaluates your data, off-site, against competitors you never see. Presentation stops being a differentiator and data completeness starts being one.
- What is the Agentic Commerce Protocol?
- It is one of the emerging open specifications for agent-initiated checkout — how an agent presents a cart, authorises payment and receives order status. It addresses the transaction layer only; discovery and comparison still run on ordinary crawl access and structured data.
- Should merchants invest in agentic commerce now?
- In the discovery, comparison and decision-support layers, yes — that work is inexpensive and overlaps with existing SEO and feed hygiene. In the transaction layer, wait: standards have not consolidated and early checkout implementations have already been reworked.
- How do I know whether agents can read my store?
- Fetch a product page without JavaScript and check for Product JSON-LD, verify your robots.txt does not block retrieval user-agents, and confirm your sitemap lists product URLs. Prefero's free scan does all three and scores the result.
Cover the layers that are already live
Discovery, comparison and decision support work today. Find out whether your store survives them.