# Agent-ready ecommerce: how to make your store work with AI agents

> An agent-ready (agent-friendly) ecommerce store is one AI agents can reach, read and operate: AI search bots and shopping agents aren't blocked, product facts are in the server HTML and structured data, key actions are real labelled controls, and ideally a protocol like UCP lets agents build carts and check out through an API.

- Source: https://is-agentic-storefront.com/guides/agent-ready-ecommerce
- Updated: 2026-10-15

## What is an agent-ready ecommerce store?

An agent-ready ecommerce store is a store that AI agents can **reach, read and operate** on behalf of a shopper. "Agent-friendly" means the same thing. In practice an agent-ready store meets four conditions:

1. **Reachable**: AI search bots and user-initiated agents get a normal `200` response, not a `403`, a bot challenge or a CAPTCHA.
2. **Readable**: product name, price, currency and availability are in the server-rendered HTML and in [Product structured data](https://developers.google.com/search/docs/appearance/structured-data/merchant-listing), so an agent that doesn't run JavaScript still sees them.
3. **Operable**: add to cart, size and color selectors, and search are real HTML controls with accessible names, so browsing agents can find and use them.
4. **Callable** (advanced): the store publishes a machine-readable commerce interface, such as a [Universal Commerce Protocol (UCP)](https://ucp.dev/latest/specification/overview/) profile, so agents can search the catalog, build a cart and start checkout through an API.

The [Agent-Friendly Ecommerce Store Scorer](https://is-agentic-storefront.com/) tests all four from the store URL alone and turns them into one score. The rest of this guide explains each part and links to a page per check.

## How do AI agents shop online?

AI agents reach ecommerce stores in three main ways. Each one needs something different from the store, which is why agent readiness has several categories rather than one.

### AI search bots and fetchers

AI assistants send their own crawlers and fetchers to stores: `OAI-SearchBot` and `ChatGPT-User` from [OpenAI](https://developers.openai.com/api/docs/bots), `Claude-SearchBot` and `Claude-User` from [Anthropic](https://support.claude.com/en/articles/8896518), `PerplexityBot` and `Perplexity-User` from [Perplexity](https://docs.perplexity.ai/guides/bots), and others. Search bots index products so they can appear in AI answers. User-initiated fetchers load a page in real time because a shopper asked about it.

These bots read raw HTML. A log study by Vercel found that GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot and PerplexityBot did not execute JavaScript ([Vercel](https://vercel.com/blog/the-rise-of-the-ai-crawler)). If your price only appears after client-side rendering, these agents don't see it. The full list of tokens is in [AI crawler and agent user agents](https://is-agentic-storefront.com/guides/ai-user-agents).

### Browsing and computer-use agents

Browsing agents such as ChatGPT agent and Atlas, Claude in Chrome, and Gemini in Chrome open your store in a real browser and click through it. Google's Lighthouse puts it plainly: "Agents rely on the accessibility tree as their primary data model" ([Chrome](https://developer.chrome.com/docs/lighthouse/agentic-browsing/accessibility-for-agents)). OpenAI asks sites to add "descriptive roles, labels, and states to interactive elements" ([OpenAI](https://help.openai.com/en/articles/12627856)). Screenshots are the fallback.

Two consequences for stores:

- A clickable `<div>` or an unlabelled icon button is invisible or ambiguous to the agent.
- CAPTCHAs are a hard stop. Google's computer-use model "MUST NEVER attempt to solve or bypass" them ([Gemini docs](https://ai.google.dev/gemini-api/docs/computer-use)).

### Protocol-based agents

Protocol-based agents skip the HTML and call an API. [UCP](https://ucp.dev/latest/specification/overview/), an open standard published at ucp.dev, lets a store publish a profile at `/.well-known/ucp` that lists its capabilities (catalog, cart, checkout, orders) and an endpoint, often an [MCP](https://modelcontextprotocol.io/) server. Some hosted commerce platforms serve this on store domains by default.

OpenAI's [Agentic Commerce Protocol (ACP)](https://developers.openai.com/commerce) works through product feeds pushed to OpenAI, so it can't be seen from a store URL. Medusa has an [official ACP tutorial](https://docs.medusajs.com/resources/how-to-tutorials/tutorials/agentic-commerce). In March 2026 OpenAI moved ChatGPT away from Instant Checkout toward discovery plus merchant checkout ([Digital Commerce 360](https://www.digitalcommerce360.com/2026/03/06/openai-shifts-checkout-plans-agentic-commerce-strategy/)). That's one reason readable product pages still matter most. See [agentic commerce protocols](https://is-agentic-storefront.com/guides/agentic-commerce-protocols) for UCP, ACP, AP2, MCP and A2A side by side.

## How is agent readiness scored?

The scorer groups 29 checks into five weighted categories. Each check is High (3), Medium (2) or Low (1) weight, and five bonus checks only add points. Full formula, grades and limits are in the [scoring methodology](https://is-agentic-storefront.com/guides/scoring-methodology).

| Category                                                                  | Weight | Question it answers                                        |
| ------------------------------------------------------------------------- | ------ | ---------------------------------------------------------- |
| [Access](#access-25-can-ai-agents-reach-your-store)                       | 25%    | Can AI agents load your pages at all?                      |
| [Product data](#product-data-25-can-ai-agents-read-your-products)         | 25%    | Can they read product facts without JavaScript?            |
| [Operability](#operability-20-can-browsing-agents-use-your-pages)         | 20%    | Can a browsing agent use add to cart, variants and search? |
| [Discovery](#discovery-15-can-ai-agents-find-your-catalog)                | 15%    | Can they find every product and fetch pages efficiently?   |
| [Agentic commerce](#agentic-commerce-15-can-ai-agents-buy-through-an-api) | 15%    | Can they search, cart and check out through an API?        |

## Access (25%): can AI agents reach your store?

Access measures whether AI agents can load your store at all. It carries the highest weight with product data because nothing else matters if an agent gets a `403`. Cloudflare has blocked AI crawlers by default on new zones since July 2025 ([Cloudflare](https://www.cloudflare.com/press/press-releases/2025/cloudflare-just-changed-how-ai-crawlers-scrape-the-internet-at-large/)), so many stores block agents without knowing it.

The scorer does not penalize blocking **training** crawlers like GPTBot or ClaudeBot. It does penalize blocking AI search bots and user-initiated agents, because that removes you from AI answers and breaks live shopping requests.

- [AI agents and AI search bots are not blocked](https://is-agentic-storefront.com/checks/agent-access) (High): your product page is fetched with each AI search and user-agent UA and compared to a browser fetch.
- [robots.txt allows AI agents and AI search bots](https://is-agentic-storefront.com/checks/robots-txt) (High): rules are evaluated per bot under [RFC 9309](https://www.rfc-editor.org/rfc/rfc9309.html).
- [No CAPTCHA on the homepage, product pages or cart](https://is-agentic-storefront.com/checks/no-captcha) (High): reCAPTCHA, hCaptcha or Turnstile on shopping pages.
- [No age gate, consent wall or geo-gate hiding content](https://is-agentic-storefront.com/checks/content-gates) (Medium): interstitials that replace the page content.

## Product data (25%): can AI agents read your products?

Product data measures whether an AI agent can read product facts, meaning name, price, currency, availability, variants and policies, straight from your HTML. OpenAI says ChatGPT shopping uses "structured metadata from first-party and third-party providers" ([OpenAI](https://help.openai.com/en/articles/11128490-shopping-with-chatgpt-search)). Google can update Merchant Center products from on-page structured data ([Google](https://support.google.com/merchants/answer/7052112)).

Agents quoting a wrong price is the worst outcome, so the scorer also checks that the JSON-LD price matches the visible one.

- [Product name, price and availability are in the server HTML](https://is-agentic-storefront.com/checks/ssr-product-facts) (High): no JavaScript needed to see them.
- [Product structured data meets Google merchant listing requirements](https://is-agentic-storefront.com/checks/product-json-ld) (High): `Product` + `Offer` with `price`, `priceCurrency` and `availability`.
- [Product variants are modeled in structured data](https://is-agentic-storefront.com/checks/product-variants) (Medium): [`ProductGroup`](https://developers.google.com/search/docs/appearance/structured-data/product-variants) with `hasVariant` and `variesBy`.
- [Return policy and shipping details are in structured data](https://is-agentic-storefront.com/checks/return-shipping-data) (Medium): [`MerchantReturnPolicy`](https://schema.org/MerchantReturnPolicy) and [`OfferShippingDetails`](https://schema.org/OfferShippingDetails).
- [OpenGraph product and price tags](https://is-agentic-storefront.com/checks/opengraph-product) (Low): `og:price:amount` / `product:price:amount` as a fallback.

## Operability (20%): can browsing agents use your pages?

Operability measures whether a browsing agent can actually use your product page: pick a size, add to cart, search, and find your policies. Because browsing agents read the accessibility tree, the test is whether your controls are real HTML elements with names. These checks overlap heavily with [WCAG 2.2](https://www.w3.org/TR/WCAG22/), so fixing them helps human shoppers too.

- [Add-to-cart is a real, named button in a form](https://is-agentic-storefront.com/checks/add-to-cart) (High): a `<button>` inside a `<form>`, not a clickable `<div>`.
- [Size/color selectors are real form controls](https://is-agentic-storefront.com/checks/variant-controls) (High): `<select>`, radio inputs, or buttons with `aria-pressed` / `aria-checked`.
- [Buttons, links and form fields have accessible names](https://is-agentic-storefront.com/checks/accessible-names) (High): no icon-only buttons without `aria-label`.
- [Site search works through a GET URL](https://is-agentic-storefront.com/checks/site-search) (High): `/search?q=shoes` returns results an agent can load directly.
- [Returns, shipping, terms, privacy and contact pages are linked sitewide](https://is-agentic-storefront.com/checks/policy-links) (Medium): agents check policies before buying.
- [Landmarks, one h1, page language and image alt text](https://is-agentic-storefront.com/checks/semantic-structure) (Low): `<main>`, `<nav>`, a single `<h1>`, `<html lang>`, `alt`.

## Discovery (15%): can AI agents find your catalog?

Discovery measures whether AI agents can enumerate your products and fetch pages cheaply. Sitemaps are the main way crawlers list a catalog ([sitemaps.org](https://www.sitemaps.org/protocol.html)), and correct status codes stop agents from indexing dead pages. Markdown versions and `llms.txt` are newer, low-weight signals that cut token cost for agents.

- [Sitemap is declared in robots.txt and lists live product URLs](https://is-agentic-storefront.com/checks/sitemap) (High): a `Sitemap:` line, and sampled product URLs return `200`.
- [Canonical URLs and correct HTTP status codes (no soft 404s)](https://is-agentic-storefront.com/checks/canonical-status) (Medium): missing products return `404`, not a `200` homepage.
- [Fast response and reasonable page weight](https://is-agentic-storefront.com/checks/performance) (Medium): time to first byte and HTML size.
- [Markdown version of pages for agents](https://is-agentic-storefront.com/checks/markdown-alternative) (Low): `Accept: text/markdown` or a `.md` alternate ([Cloudflare](https://developers.cloudflare.com/fundamentals/reference/markdown-for-agents)).
- [/llms.txt or /agents.md guides agents to your catalog](https://is-agentic-storefront.com/checks/llms-txt) (Low): a curated Markdown map ([llmstxt.org](https://llmstxt.org/)).

## Agentic commerce (15%): can AI agents buy through an API?

Agentic commerce measures whether your store exposes an interface agents can call to search, build a cart and check out, instead of clicking through pages. UCP is the only commerce protocol with a public, crawlable discovery file, so it anchors this category. Today the category mostly tracks whether a store runs on a platform that ships UCP by default, so its weight is held at 15%.

- [Universal Commerce Protocol (UCP) profile](https://is-agentic-storefront.com/checks/ucp-profile) (High): a valid `/.well-known/ucp`, graded by capability depth.
- [Live agent endpoint (MCP/REST) from the UCP profile](https://is-agentic-storefront.com/checks/agent-endpoint) (High): the endpoint answers, for example, an MCP `tools/list` call.
- [Cart/checkout permalinks agents can hand to shoppers](https://is-agentic-storefront.com/checks/checkout-permalinks) (Medium): a URL that builds a cart and redirects to checkout ([UCP permalink](https://ucp.dev/2026-08-25/specification/permalink/)).
- [OAuth discovery metadata for agent sign-in](https://is-agentic-storefront.com/checks/oauth-discovery) (Medium): [RFC 9728](https://datatracker.ietf.org/doc/html/rfc9728) protected resource metadata.

Bonus signals (Low, only add points when found):

- [WebMCP tools for in-browser agents](https://is-agentic-storefront.com/checks/bonus-signals#webmcp)
- [A2A agent card](https://is-agentic-storefront.com/checks/bonus-signals#a2a-agent-card)
- [AI Catalog (MCP server cards)](https://is-agentic-storefront.com/checks/bonus-signals#ai-catalog)
- [API catalog (RFC 9727)](https://is-agentic-storefront.com/checks/bonus-signals#api-catalog)
- [UCP profile advertised with rel="ucp"](https://is-agentic-storefront.com/checks/bonus-signals#ucp-link)

## Where should I start making my store agent-ready?

Work through fixes in this order. It follows the weights, so each step moves the score as much as possible for the effort.

1. **Unblock AI search bots and user agents.** Check your WAF or CDN bot settings and your [robots.txt](https://is-agentic-storefront.com/checks/robots-txt). Keep training crawlers blocked if you want; that isn't penalized.
2. **Remove CAPTCHAs from home, product and cart pages.** Keep them on login and contact forms if you need them.
3. **Server-render product facts.** Name, price, currency and availability belong in the HTML response, not only in client-side JavaScript.
4. **Add complete Product JSON-LD.** Include `offers` with `price`, `priceCurrency` and `availability`, and make sure it matches the visible price.
5. **Make add to cart and variant selectors real controls.** Use a `<button>` in a `<form>` and native `<select>` or radio inputs.
6. **Declare a sitemap in robots.txt** and make sure missing product URLs return `404`.
7. **Expose search as a GET URL** and link your returns, shipping and contact pages in the footer.
8. **Add return and shipping structured data.** Only 3 of 31 stores in our calibration set had it, so it's an easy way to stand out.
9. **Then add agentic commerce:** a UCP profile, a live endpoint, and cart permalinks.

[Scan your store](https://is-agentic-storefront.com/) to see which of these already pass.

## How do I make a store agent-ready on my platform?

The fixes depend on your platform. Some hosted platforms ship UCP and `llms.txt` by default but depend on the theme for JSON-LD. Headless stores control everything but start with little. Platform guides:

- [Medusa agent readiness](https://is-agentic-storefront.com/guides/medusa-agent-readiness): what the Medusa starter storefront gives agents by default, what to add, and which agentic commerce protocols Medusa supports.

## Related guides

- [AI crawler and agent user agents: the 2026 reference for stores](https://is-agentic-storefront.com/guides/ai-user-agents)
- [How the agent-readiness score works](https://is-agentic-storefront.com/guides/scoring-methodology)
- [Agent readiness vs SEO: what's different for ecommerce](https://is-agentic-storefront.com/guides/agent-readiness-vs-seo)
- [Agentic commerce protocols](https://is-agentic-storefront.com/guides/agentic-commerce-protocols)
- [Glossary of agent-readiness terms](https://is-agentic-storefront.com/glossary)

## Frequently asked questions

### What does agent-ready mean for an ecommerce store?

An agent-ready store lets AI agents do what a shopper does: load pages without being blocked, read product names, prices and availability without running JavaScript, use add-to-cart, variant and search controls, and, on the most advanced stores, find a commerce API such as a UCP profile at /.well-known/ucp.

### Is agent readiness the same as SEO?

No. They overlap on sitemaps, structured data, server rendering and canonical URLs, but agent readiness also covers robots.txt rules for AI bots, bot-protection behavior, whether browsing agents can operate the page, and commerce protocols like UCP and MCP that search engines don't use.

### Should I block AI crawlers in robots.txt?

Blocking AI training crawlers such as GPTBot, ClaudeBot or Google-Extended is a business choice and doesn't hurt agent readiness. Blocking AI search bots (OAI-SearchBot, Claude-SearchBot, PerplexityBot) or user-initiated agents (ChatGPT-User, Claude-User) removes your products from AI answers and breaks live shopping requests.

### What should I fix first to make my store agent-ready?

Start with access: make sure AI search bots and user agents get a 200 with real content and no CAPTCHA. Then put product name, price, currency and availability in the server HTML and in Product JSON-LD. Those two steps cover the highest-weighted checks.

### Do I need UCP or MCP to be agent-ready?

Not to start. Agentic commerce is 15% of the score. Most agents today still read your pages and send shoppers to your checkout, so access, product data and operable pages matter more. UCP becomes worth adding once the basics pass.

## Run your store on a platform built for agentic commerce

Medusa is open-source, headless commerce with a full Store API agents can call and storefronts you control end to end.

### Deploy on Medusa Cloud

Open-source commerce with a full Store API, hosted and scaled for you. Your storefront, your data model.

- [Deploy on Medusa Cloud](https://cloud.medusajs.com?utm_source=agent-friendly-scorer&utm_medium=referral&utm_campaign=markdown-cta&utm_content=%2Fguides%2Fagent-ready-ecommerce)
- [Medusa Cloud docs](https://docs.medusajs.com/cloud?utm_source=agent-friendly-scorer&utm_medium=referral&utm_campaign=markdown-cta&utm_content=%2Fguides%2Fagent-ready-ecommerce)

### Storefront skills for coding agents

Teach Claude Code, Cursor or Codex to build storefronts with the patterns these checks look for.

```bash
npx skills add medusajs/medusa-agent-skills
```

[View the skill on GitHub](https://github.com/medusajs/medusa-agent-skills/tree/main/plugins/ecommerce-storefront/skills/storefront-best-practices)

---

Scan your store for free at https://is-agentic-storefront.com/