Ottom

Learn / Start here

What is agentic commerce?

Agentic commerce is shopping carried out by an AI agent acting on a person’s behalf: the shopper states an intent, and software finds candidate products, compares them, and completes the purchase. The shopper may never load the storefront, but as of 2026 the purchase itself often completes on the merchant’s own checkout.

That last sentence is the part worth sitting with. For twenty years, ecommerce has optimised for a human eye arriving on a page. Agentic commerce optimises for a program reading data. The two want different things, and a store can be excellent at one while failing entirely at the other.

What the agent actually does

A shopping agent works through four stages, and it can fail at any of them:

  1. Discover. Find your store and your catalogue at all. Robots rules, sitemaps, product feeds and structured data decide whether you exist to an agent.
  2. Evaluate. Work out what you sell, for how much, whether it is in stock, and what your terms are. This requires text, not layout.
  3. Compare. Weigh you against an alternative. If your shipping cost only appears at checkout, the agent has to guess your true total, and a guess is easy to lose.
  4. Buy. Get a real product to the payment step. This is where variant pickers, popups, forced account creation and bot walls end the journey.
The four stages of an agent purchase attempt A vertical sequence: discover, then evaluate, then compare, then buy. An agent can stall at any one of the four, and a store that fails an early stage is never measured on the later ones. 1 Discover Find your store and your catalogue at all 2 Evaluate Work out what you sell, for how much, and whether it is in stock 3 Compare Weigh you against an alternative 4 Buy Get a real product to the payment step
An agent can stall at any of the four stages. A store that fails discovery is never measured on checkout.

A store that fails at discovery never gets measured on the rest. A store that passes discovery and evaluation but fails at checkout is the more common and more expensive case: the agent chose you, then could not complete.

Where the purchase actually completes

This changed during 2026, and it changed in the merchant’s favour. Retail Dive reported on 25 March 2026 that OpenAI had announced the day before that it was retiring the first version of Instant Checkout, roughly six months after launching it on 29 September 2025. Purchases that had been completing inside ChatGPT moved back to merchant sites.

Retail Dive carried OpenAI’s statement: "the initial version of Instant Checkout did not offer the level of flexibility that we aspire to provide, so we’re allowing merchants to use their own checkout experiences while we focus our efforts on product discovery."

Shopify’s help centre describes the mechanics: ChatGPT users "complete their purchase on your online store checkout in a ChatGPT in-app browser, or in a new tab when customers use ChatGPT web". Your checkout is the checkout. Anything that stalls in it now stalls a sale that an assistant had already sent you.

The other pattern did not go away. The same Shopify documentation says shoppers using Google AI Mode, Gemini or Microsoft Copilot complete their purchase "in the AI channel’s Shopify-powered direct checkout without having to leave their chat conversation". A single store can be selling under both models at the same time.

Why this is arriving now

The protocols landed in quick succession. Google published the Agent Payments Protocol on 16 September 2025, saying it was built with more than sixty collaborating organisations. OpenAI and Stripe published the first version of the Agentic Commerce Protocol on 29 September 2025 under Apache 2.0, according to the protocol’s own repository. Google announced the Universal Commerce Protocol at NRF on 11 January 2026, built with Shopify, Etsy, Wayfair, Target and Walmart, and said on its company blog that more than twenty further companies had endorsed it.

Then the platforms turned it on. Shopify’s documentation says agentic storefronts are active by default for eligible stores, which means a merchant can already be taking agent traffic without ever having opted into anything.

You do not have to believe agentic commerce will be the majority of retail to care about it. You only have to believe it will be non-zero, and that being invisible to it is a choice you would rather make deliberately.

What makes a store agent-ready

Almost none of it is exotic:

  • A product feed and a sitemap, so an agent can read your catalogue instead of scraping it.
  • schema.org Product markup with price, availability and an identifier such as a GTIN.
  • Policies published as text, not as images. Shopify lists published store policies among the requirements for selling through agentic storefronts.
  • Variant selection that responds to a click and a keyboard, not only a hover.
  • A checkout that survives a mobile in-app browser, because that is where a ChatGPT purchase now lands.
  • Bot protection that does not challenge identifiable, well-behaved agents.
What a shopper sees against what an agent reads The same product page. A person sees an image, a price rendered by script, and a policy set in a graphic. An agent reads only the text and structured data, so the price and the policy are missing from its view. WHAT A SHOPPER SEES product photo Trail Jacket £129.00 In stock, ships tomorrow Free returns for 30 days WHAT AN AGENT READS image: no alt text Trail Jacket price: not found availability: not stated returns: image only
The same product page, twice. Your store looks fine in a browser, which is why this is invisible to you.

The uncomfortable part is that these are invisible to you. Your store looks fine in a browser. The only way to know how it looks to an agent is to send one.

Questions

Is agentic commerce the same as an AI chatbot on my store?

No. A chatbot helps a human who is already on your site. An agent acts for someone who may never visit it. It reads your data, decides whether you are the right store, and transacts. The chatbot is a feature of your storefront; the agent is a new kind of customer arriving at it.

Do I need to do anything differently?

Mostly you need your existing product data to be readable without a browser. An agent cannot hover, cannot wait for a carousel, and will not guess your returns policy from a graphic. Most of what makes a store agent-ready is unglamorous data hygiene.

Where does the purchase actually happen?

It depends on the assistant. Shopify’s help centre says ChatGPT shoppers "complete their purchase on your online store checkout in a ChatGPT in-app browser, or in a new tab when customers use ChatGPT web", so for ChatGPT the answer is your checkout. The same documentation says shoppers using Google AI Mode, Gemini or Microsoft Copilot complete "in the AI channel’s Shopify-powered direct checkout without having to leave their chat conversation". Both patterns are live at once, so a store needs its data readable and its checkout operable regardless of which one sends the order.

Scan your store