For AI agents: a documentation index is available at /llms.txt
ACO · Agentic commerce
ACO (agentic commerce optimization)
ACO (agentic commerce optimization) gets your products found, compared and bought by AI agents shopping for a buyer, with your store still the seller. It is the last step after SEO, GEO and AXO (agent experience optimization), and it starts with product data that agents can read and trust.
- Product
- Standing desk, oak
- Price
- USD 590.00
- Availability
- In stock
- Feed and page agree
- Yes
- ACPAgentic Commerce Protocol readinessIn progress
- UCPUniversal Commerce Protocol readinessIn progress
01
What ACO covers
ACO makes your catalogue something an agent can work with: complete product data, prices, stock and policies that agree everywhere, and support for the checkout protocols agents use. Agents compare stores on facts they can read, so a product they cannot read, or one whose price differs between page and feed, drops out of the comparison.
The term is new. Commerce vendors started using it in 2025 and there is no settled definition, so here is ours. ACO has three parts:
- Product data agents can read: titles, attributes, variants, prices, stock, shipping and returns, in the HTML and in structured data (labels that tell machines what a page is).
- Feeds that agree with your pages, so an agent never quotes a stale price.
- Readiness for agent checkout through the Universal Commerce Protocol (UCP) and the Agentic Commerce Protocol (ACP), the two standards agents use to place an order.
02
Why agentic commerce matters this year
AI shopping is still a small channel, but it is growing fast and its visitors convert well. Adobe Analytics found AI-referred traffic to US retail sites grew 138% year on year in May 2026, and those visits converted 54% better than traffic from other sources.
Agents now act as well as advise. Meta's Muse, launched in September 2026, can open a browser, fill in forms and buy, asking the user to approve a purchase first. When an agent shortlists products for a buyer, the stores it can read are the ones it puts forward.
03
How agent checkout works today
Agent checkout exists, but most AI shopping still ends on the store's own site. Google announced UCP in January 2026, built with Shopify, Etsy, Wayfair, Target and Walmart, and UCP Checker counted more than 19,000 stores publishing a valid UCP profile in October 2026.
OpenAI and Stripe launched ACP with Instant Checkout in ChatGPT in September 2025. In March 2026 OpenAI moved those purchases into retailers' own apps inside ChatGPT, after Walmart reported that in-chat checkout converted worse than sending shoppers to its site. ACP itself continues. So in our judgement, clean product data pays off now under every protocol, while a custom checkout integration can wait until your platform supports one.
04
What to fix first
Start with the product facts an agent compares, and leave checkout engineering for later. Each step builds on the one before:
| Step | What to do | Why it comes first |
|---|---|---|
| 1. Product facts in the HTML | Put prices, variants, stock and return rules in the page itself, not only in scripts | AI fetchers skip JavaScript, so facts loaded by scripts go unseen |
| 2. Pages and feeds agree | Match titles, prices and stock across pages, feeds and structured data | A contradiction makes an agent distrust or skip the product |
| 3. Agent-ready site | Clear form labels, llms.txt and AI crawler rules, as on agent-ready websites | Agents that can read and use your site can finish a purchase there |
| 4. Protocol readiness | Ask your commerce platform how it supports UCP and ACP | Most stores get checkout support from their platform, not custom code |
Say a buyer asks an assistant for a waterproof jacket under a set price with free returns. The assistant checks feeds, page data and return policies across several stores. If your stock status loads only in a script, or your feed price differs from the page, your jacket is less likely to make the shortlist.
05
How we measure it
We check your store before and after each fix. Each check on your product pages, feeds, structured data and site files is scored Pass, Partial, Fail or Not applicable, and a missing optional file is Not applicable, never a fail.
For AI shopping answers in ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews and AI Mode, Named means your brand appears in the answer and Cited means one of your pages is a source. Our browser agent tests key tasks but never submits a form or buys anything. We do not guarantee picks or sales. The work starts with a free GEO and AXO audit.
Questions
Questions buyers ask
What is agentic commerce?
Agentic commerce is shopping done by an AI agent for a person. The agent researches products, compares prices, stock and return terms across stores, and can place the order once the person approves. It reads your product data, feeds and structured data, and those decide whether your products make its shortlist.
What does ACO stand for, and how is it different from AXO?
ACO stands for agentic commerce optimization. AXO (agent experience optimization) makes any website readable and usable by agents, for tasks like finding pricing or booking a demo. ACO applies that to selling products: the agent has to find, compare and buy them, so product data, feeds and checkout protocols matter most.
Are AI agents actually buying products yet?
Yes, but at small volume. Google's UCP supports checkout inside its AI experiences, and Meta's Muse can buy for a user after asking for approval. OpenAI moved ChatGPT purchases into retailers' own apps in March 2026. Most agent shopping still ends on the store's site, which is why product data comes first.
What are UCP and ACP?
They are open standards for how AI agents check out from a store. Google announced the Universal Commerce Protocol (UCP) in January 2026 with large retailers. OpenAI and Stripe launched the Agentic Commerce Protocol (ACP) in September 2025. Your commerce platform usually decides which one your store supports, so ask your vendor before building anything custom.
Why do AI answers show the wrong price or stock for our products?
Usually because the price or stock loads through JavaScript, which AI fetchers skip, or because your feed and page disagree. The AI then quotes an older page or another seller. Putting product facts in the HTML and structured data, and keeping feeds in sync, fixes most cases.
Will your test agent place orders in our store?
No. Our browser agent tests key tasks, such as finding a product's price or reaching the cart, and records the step where it stopped. It never submits a form, places an order or holds stock. Your team approves every change we propose before it ships.