Agentic commerce is online shopping carried out by an AI agent on a person’s behalf — the agent browses, compares products across different stores, answers the shopper’s follow-up questions and, increasingly, completes the purchase itself. Instead of a person opening ten tabs and reading product pages, they describe what they want to an assistant like ChatGPT, Perplexity or Gemini, and the assistant does the legwork.
The term is everywhere in 2026, and most of what is written about it is either a vendor press release or a developer specification. This guide does neither. It explains, in plain English, what agentic commerce actually is, how it works layer by layer, who is building the infrastructure right now, and what — if anything — you need to do about it as a merchant. We will stick to what is built and shipping, and flag clearly where things are still moving.
Agentic commerce, defined
Agentic commerce is a shift in who does the shopping. For twenty-five years the model has been the same: a person visits your store, reads your pages, and decides. Agentic commerce inserts an AI agent between the shopper and the store. The person states an intent — "find me a waterproof jacket under £150 that packs down small" — and the agent carries out the research, the comparison and, where the store supports it, the checkout.
That is different from a chatbot on your website answering questions, and different again from a recommendation engine suggesting products. The defining feature is autonomy across stores: the agent is not working for you, it is working for the shopper, and it will weigh you against competitors before it recommends anything. Being visible to that agent, and being the option it picks, is the whole game.
A useful analogy is a personal shopper. You would not hand a personal shopper a list of websites to visit; you would describe what you want and trust them to know the market, ask the right questions and come back with the best option. An AI agent plays that role at scale, for millions of shoppers at once. The stores that win a personal shopper’s recommendation are the ones whose products they understand well and whose questions they can answer on the spot — and agentic commerce works exactly the same way, except the "personal shopper" is reading structured data rather than remembering a relationship.
How agentic commerce actually works
It helps to think of agentic commerce as three layers stacked on top of each other: the agent that talks to the shopper, the protocol that lets that agent talk to stores, and the checkout layer that lets the purchase actually happen. Each is being built by different companies, and each has to work for the whole thing to function.
The agent layer: ChatGPT, Perplexity and Gemini
The top layer is the assistant the shopper actually talks to. ChatGPT can now search for products, compare them and — through Instant Checkout — let the user buy without leaving the conversation. Perplexity runs a shopping experience with its own merchant programme. Google is weaving shopping into Gemini and its AI Overviews. These agents are the shop window of agentic commerce: if your products are not legible to them, none of the layers below matter.
The protocol layer: the Agentic Commerce Protocol
For an agent to buy from a store it has never seen, there has to be a shared language for "here is a product, here is its price and availability, here is how to place an order". That is what the Agentic Commerce Protocol (ACP) is — an open standard, developed by OpenAI with Stripe, that defines how an agent discovers products and submits an order to a merchant. Google is building a parallel stack with its Agent Payments Protocol (AP2) and a Universal Cart. These protocols are the plumbing: unglamorous, invisible to shoppers, and the reason an agent can transact with thousands of stores without a bespoke integration for each one.
The checkout and payment layer: Stripe and Shopify Instant Checkout
The final layer is money changing hands safely. This is where Stripe agentic commerce comes in: Stripe’s Shared Payment Tokens let an agent complete a payment on the shopper’s behalf without ever handling raw card details, using a scoped, single-use mandate. On Shopify, this surfaces as Instant Checkout — the order is placed inside the AI conversation, but the merchant remains the merchant of record, keeping the customer relationship, the data and the fulfilment. The shopper gets a one-step purchase; you get an order that looks, from your side of the counter, much like any other.
Who’s building this right now
This is a fast-moving space, and the specifics below will keep shifting — treat them as a snapshot, not a permanent map. That said, the shape of it is already clear, and it is being built by the companies you would expect.
Shopify has moved fastest for merchants. Shopify Catalog syndicates product data out to AI surfaces, Instant Checkout lets agents complete orders, and the platform has rolled these "agentic storefront" capabilities out broadly rather than as a niche beta. For a Shopify merchant, a lot of the distribution is now switched on by default — which is exactly why being chosen, not just listed, is the thing that matters. Our companion guide, Shopify agentic commerce: what merchants need to know, covers the platform specifics.
OpenAI owns the highest-traffic agent (ChatGPT) and co-authored the Agentic Commerce Protocol with Stripe, giving it both the shop window and the plumbing. Instant Checkout in ChatGPT is the most visible consumer-facing example of agentic commerce shipping today.
Stripe is quietly central. Its Shared Payment Tokens are the mechanism that lets an agent pay without the merchant or the model touching card data, and Stripe’s involvement in ACP means a large share of online checkout infrastructure is being made agent-aware at once. Where Stripe goes, a lot of commerce follows.
If you want to go past the summary, the primary sources are worth reading directly: Shopify’s developer documentation covers Catalog and Instant Checkout, and OpenAI and Stripe have both published the Agentic Commerce Protocol specification. Because the space is evolving quickly, those will always be more current than any single guide.
Agentic commerce examples you can see today
The theory is easier to grasp with concrete agentic commerce examples that already exist:
- ChatGPT Instant Checkout. A shopper asks ChatGPT for a product, it returns options with prices and details, and for supported merchants a buy step completes the order inside the chat. This is the most visible consumer-facing example shipping today.
- Perplexity shopping. Perplexity surfaces products in its answers and runs a merchant programme with a "Buy with Pro" checkout, so a research query can end in a purchase without leaving the assistant.
- Google’s AI surfaces. Google is weaving products into AI Overviews and Gemini, with its Agent Payments Protocol and Universal Cart handling the transaction side across merchants.
What ties the examples together is the pattern, not the brand: an assistant does the research and comparison, and the purchase closes inside the conversation. Whichever agent your customers use, the demand it places on your store is the same — be legible, be answerable, be transactable.
What agentic commerce means for merchants — not just theory
Strip away the hype and there are three practical questions every merchant should ask. None of them require you to be a developer to understand.
1. Discoverability — can an agent find and understand your products? An agent does not browse your beautifully designed category page; it reads structured data. If your product titles are vague, your attributes thin and your catalogue poorly organised, the agent cannot tell what you sell, and it recommends a competitor it does understand. Discoverability in agentic commerce is a data problem, not a design problem.
2. Structured data — is your product information machine-readable? This is the single biggest lever. Complete Product and Offer schema, accurate prices and availability, real specifications held as structured attributes rather than buried in a paragraph — these are what let an agent compare your item against three others on the criteria the shopper actually cares about. Thin data does not just rank badly; it makes you invisible at the moment of comparison.
3. Trust and checkout — can an agent actually complete a purchase? Visibility is worthless if the transaction cannot close. On a supported platform like Shopify, being connected to the agentic checkout path is what turns "the agent mentioned you" into "the agent bought from you". Off-platform, it means wiring up the protocols directly. Either way, the store that an agent can transact with beats the store it can only describe.
Is agentic commerce relevant to my store yet?
Honest answer: it is early, but it is moving faster than most channels did. Agentic commerce is not yet a large share of most merchants’ revenue, and anyone telling you to drop everything is overselling. But two things make it worth attention now rather than later. First, the work that prepares you for it — clean product data, complete schema, a knowledge base agents can quote — is the same work that improves your ordinary SEO and conversion, so it is rarely wasted effort. Second, the field is unusually open: because most merchants have done nothing, being early is cheap and being late may not be.
A sensible posture for most stores is: get the fundamentals right now, because they pay off across every channel; connect to the agentic infrastructure your platform offers; and measure what comes through it before deciding how much further to invest. That is a world away from a gold rush, and it is the honest version of the advice.
Frequently asked questions
What is the Agentic Commerce Protocol (ACP)?
The Agentic Commerce Protocol is an open standard, developed by OpenAI with Stripe, that defines how an AI agent discovers a merchant’s products and submits an order on a shopper’s behalf. It is the shared language that lets an agent transact with many stores without a custom integration for each one. Google is building a parallel stack with its Agent Payments Protocol (AP2) and a Universal Cart.
Is agentic commerce the same as AI shopping?
They overlap but are not identical. "AI shopping" usually means an assistant helping a person research and compare products. Agentic commerce goes one step further: the agent can also act — placing the order itself through a checkout protocol. All agentic commerce is AI shopping, but not all AI shopping completes the purchase.
Do I need to do anything to prepare my Shopify store?
The highest-value preparation is getting your product data and schema complete and accurate, and giving agents a knowledge base they can quote for shipping, returns and sizing questions. On Shopify much of the distribution is switched on by default, so the work is less about connecting and more about being the store the agent chooses. Our Shopify-specific guide walks through the specifics.
Is agentic commerce only for Shopify stores?
No. Shopify has made it easiest, but the underlying protocols — ACP, Stripe’s payment tokens, Google’s AP2 and Universal Cart — are platform-agnostic. WooCommerce, Magento, custom and headless stores can all participate; they simply have to wire up the protocols directly rather than relying on the platform to do it for them.
Wondering where your store stands? Our Agentic Storefront Optimization service opens with a free Agentic Readiness Audit — we score your catalogue data, knowledge coverage and checkout readiness, and show you how AI agents answer buying questions in your category today.