Until now, AI answered. Tomorrow, it will act: an agent that searches, compares several products, fills a cart, and sometimes places the order — on your behalf. This is agentic commerce. Still emerging, but it raises a simple question: is your store ready to be handled by a machine, not just read by a human?
What is agentic commerce?
An AI shopping agent is an autonomous assistant that carries a commercial task end to end: "find me the best fragrance-free SPF 50 sunscreen under €30, and add it to the cart". It doesn't just suggest: it evaluates, compares, and may go all the way to action. There are three levels of autonomy:
- Recommend: the agent proposes products (close to today's answer engine).
- Compare: it weighs several options on precise criteria and decides.
- Transact: it fills the cart, or even triggers the purchase.
From answer engine to agent that acts
The difference from classic GEO is one of intensity. An answer engine cites a source; an agent has to decide and sometimes commit a transaction. The machine-readability bar rises a notch: being findable is no longer enough, you must be usable all the way through — price, stock, terms, identifiers.
What an agent checks before buying
Decisive facts, structured
As with GEO, the agent needs the deciding attributes: SPF, compatibility, size, material. In structured data, not in an image. Without them, it can't verify your product meets the request — it moves on to the next one.
A reliable price and availability, in real time
An agent about to buy doesn't tolerate approximations: a stale price or wrong stock is a deal-breaker. The freshness and accuracy of your feed become critical — far more than for a simple human display.
Clear terms (shipping, returns)
Shipping cost, delivery times, return policy: an agent that compares takes them into account. Exposing these terms in a machine-readable way can make the difference between being kept or dropped at the final step.
{
"@context": "https://schema.org",
"@type": "Offer",
"price": "27.90",
"priceCurrency": "EUR",
"availability": "https://schema.org/InStock",
"shippingDetails": {
"@type": "OfferShippingDetails",
"shippingRate": { "@type": "MonetaryAmount", "value": "0", "currency": "EUR" },
"deliveryTime": { "@type": "ShippingDeliveryTime", "transitTime": "P2D" }
},
"hasMerchantReturnPolicy": {
"@type": "MerchantReturnPolicy",
"returnPolicyCategory": "https://schema.org/MerchantReturnFiniteReturnWindow",
"merchantReturnDays": 30
}
}Here, the agent reads everything it needs to decide: price, availability, free shipping, 2-day delivery, 30-day return. No ambiguity to resolve, no reason to drop you.
The state of play: early, and shifting
Let's be clear-eyed: agentic commerce is in its early days. Agent protocols, agentic-payment standards and checkout integrations are still taking shape, and all of it will evolve. There's no fixed recipe. But — and this is the good news — the prerequisites are exactly the GEO fundamentals: structured data, clean feed, consistency. So preparing today isn't a gamble: it's useful work no matter what.
Getting ready today (without a risky bet)
- Structure your products (
schema.org/Product, decisive attributes). - Keep a clean, up-to-date feed: accurate price, stock, availability.
- Add identifiers (GTIN) so you can be matched.
- Expose your terms (shipping, returns) in a structured way.
- Remove inconsistencies between page, feed and markup.
- Measure your presence in AI answers to track your progress.
A human forgives. An agent moves on.