AI for ecommerce
What is agentic commerce?
How AI assistants find, compare and buy products for shoppers today, and what your store needs so they pick you.
Start for freeAgentic commerce is shopping where an AI agent does the buying work for a person. The agent finds products, compares price, stock and return policies, and checks out once the shopper says yes. Today it happens inside ChatGPT, Google's AI Mode and other AI chats, where software reads your product data, not your design. Replo builds store pages with prices and stock in the page itself, so people and agents read the same facts.
Etsy sellers were among the first shops to sell this way. Etsy launched in 2005 as a market for handmade and vintage goods. By the end of 2024, it connected 8 million sellers with 96 million buyers.
On September 29, 2025, OpenAI let U.S. shoppers buy from U.S. Etsy sellers right inside ChatGPT. In OpenAI's demo, a shopper asks for handmade dinnerware under $100 and taps Buy on an $85 set from the shop BlancPottery. Etsy confirms the order in the chat, and the seller ships it as usual.

What does agentic commerce mean?
Agentic commerce means the buyer is software that follows a person's instructions. The shopper sets the goal and the limits: budget, size, brand and delivery date. The agent does the searching and comparing, and more and more often, the checkout.
Draw the line between advising and acting. A recommendation engine suggests a product, a chatbot answers a question, and an agent picks and buys.
A person weighs the photos, the headline and the reviews. An agent weighs price, stock, delivery date, the return window, and whether your data agrees with itself.

How does agentic commerce work?
Every agent purchase runs through four steps:
- The shopper states the goal, such as "waterproof hiking boots, size 8, under $150, here by Friday."
- The agent reads product data from several stores: price, stock, shipping and returns.
- The agent picks a product and builds a cart.
- A payment flow made for agents runs the payment. A token or virtual card stands in for the shopper's real card.

Store owners often miss two details. First, the agent brings bigger purchases back to the person for one approval tap, as IBM described in January 2026.
Second, the agent may not see what a shopper sees. Some pages show the price or stock only after a script runs. The agent may then read nothing and move on to a store it can read.
What does agentic commerce look like today?
Shopping in ChatGPT
ChatGPT shows products with photos and prices and builds comparison tables from product data. Etsy sellers got Instant Checkout first, and at launch it took one item per order. Today OpenAI's merchant page says ChatGPT sends shoppers to the store's own site to buy by default. Stores on Shopify or Etsy are already part of it, with no extra setup.

Google AI Mode and "Buy for me"
Google's AI Mode answers a shopping question with picked products and lets you compare them. Its agentic checkout watches a price for you. When the price drops to your target, it buys with Google Pay on the store's own site.

Which agentic commerce platforms and protocols matter?
The AI chats, the payment companies and the big marketplaces decide which standard wins. Your platform and payment processor build to these standards, so you don't. Wikipedia's roundup lists four worth knowing:
- Agentic Commerce Protocol (ACP), from OpenAI and Stripe. It covers how agents read products, build carts and check out. In ChatGPT, it powers product discovery today.
- Universal Commerce Protocol (UCP), which Google announced in January 2026 as an open standard between agents and stores.
- Agent Payments Protocol (AP2), which Google announced in September 2025. It keeps signed records of what the shopper approved, for cards, bank transfers and stablecoins.
- Agentic Commerce Trust Protocol, from Alipay, first used by Qwen with Taobao.
Every one of them reads the same thing underneath: product data that is structured, current and true. That part is your job.
What do agents need from your store?

Agents need seven things, listed in the order they most often break:
- Structured product data: price, brand, color, size and stock in schema markup.
- Pages that arrive built. Most AI crawlers read only the raw HTML. Replo pages are server-rendered, so copy, prices and product details are there before any script runs.
- Prices and stock that are true right now. An agent that quoted one price and got charged another marks your store as unreliable.
- Policies in plain text: returns, shipping times and warranty on a page anyone can read, not in an image or a PDF.
- One way to describe variants, so "the yellow dress in size 8" means the same item on every page and feed.
- Speed. Agents comparing dozens of stores skip slow pages, and so do people.
- The same facts everywhere: your ad, email and landing page should agree.
We checked the product data on nine well-known brands' product pages. Seven had product data with a price or stock status. None put a return policy or shipping time in it.
That's the cheapest gap to close. Most of this list is SEO work, so the same fixes help a page rank and help an agent buy from it.

Is agentic commerce here yet?

Shoppers already use AI to find products, but agents that buy on their own are early. IBM reported in January 2026 that 45% of consumers use AI for part of the buying journey. Yet OpenAI's merchant page now sends shoppers to the store's site to buy by default.
Search shows the same split. "Agentic commerce" drew 5,400 Google searches a month against 121 inside AI assistants (DataForSEO, US, October 2026).
Agents suit the predictable buys: refills, subscriptions, reorders and price tracking. Even OpenAI's Etsy demo bought a single dinnerware set, because Instant Checkout started with one item per order. For bigger buys, a person still looks at the product and wants it. Paid traffic still lands on a page, and that page still decides whether the click turns into a sale.
Where should you start?
- See what a machine sees. Load one product page with JavaScript turned off. Whatever is missing there is missing for agents too.
- Fix the product data so it matches the page.
- Keep price and stock in one place that feeds every page and feed.
- Write returns and shipping as text, short and dated, on a page crawlers can reach.
- Ask your platform and payment processor which agent checkout standards they support, instead of building one.
- Keep shipping and testing pages. The number of pages you launch and improve still moves revenue most.

Replo builds pages from your Shopify catalog and serves them server-rendered on your own domain. Shoppers, crawlers and agents all read the same price and stock. That shared data is the difference between Replo and an AI store builder that generates a storefront once. See what Replo builds on your catalog.
Frequently asked questions
- Can you give an example of agentic commerce?
- Yes. In OpenAI's demo, a shopper asks ChatGPT for handmade dinnerware under $100. ChatGPT shows an $85 set from the Etsy shop BlancPottery. The shopper taps Buy, confirms shipping and payment, and Etsy confirms the order in the chat.
- What does agentic selling mean?
- Agentic selling means your products get sold through an AI agent that shops for a person. The agent reads your price, stock, shipping and returns, compares you with other stores, and buys if you fit. You still ship the order and handle returns.
- Is agentic commerce the future?
- Part of it is here now. AI chats already help people find and compare products, and IBM reported in January 2026 that 45% of consumers use AI for part of the buying journey. Agents that buy on their own are still early, so plan for agents reading your store first.
- Do I need to support a protocol like ACP or AP2 to sell to agents?
- Almost certainly not by hand. Platforms, payment processors and marketplaces build these standards, and your store gets them through whatever runs your checkout. Ask your platform and processor what they support. Spend your own time on the product data every protocol reads.
- Will agentic commerce kill landing pages?
- No. Agents suit predictable buys like refills, subscriptions and reorders. For bigger buys, a person still looks at the product and decides. Your paid ads still send people to a page, and for most stores that page is still the change that moves revenue most.



