The term has no standard meaning: what Cloudflare, specification.website, Shopify and Stripe say, how Storewalk (ours) builds its Ready score, and why other tools’ scores cannot be compared with it.
Agent readiness is how well an AI agent can use your store on a shopper's behalf: find your products, read prices, stock and terms it can trust, get through checkout, and reach you after the order. It is measured by tools that read your store the way an agent would and report what they found, each on its own scale.
That definition is ours. The term has no standard meaning, and the companies that use it either word it differently or do not define it at all. Below we set out what each publisher says, how Storewalk measures it, and how other tools do. Storewalk is ours: we build it, and this blog is part of its site. We describe it from our own code and copy, and every other tool from its own pages. Where we explain a term in plain words, the explanation is ours.
If you are new to the series, the hub article sets out the five walls between an agent and a finished purchase, and what is agentic commerce? covers the shopping side of the same idea.
01Is there a standard definition of agent readiness?
No. No standards body defines it, and the publishers who use it say different things.
Cloudflare launched an Agent Readiness score in a blog post published on 17 April 2026 and last modified on 15 July 2026. The post gives no one-line definition. Its nearest wording is its own summary, which says the score "can help site owners understand how well their websites support AI agents." The score is built on four dimensions: Discoverability, Content, Bot Access Control and Capabilities. The scanner also looks for "agentic commerce standards including x402, Universal Commerce Protocol, and Agentic Commerce Protocol", but says "these do not currently count towards the score."
specification.website, a site whose pages do not name a publisher, is the one source we found with a sentence that starts "Agent readiness is". It reads: "Agent readiness is the set of choices that make a site legible to AI agents and LLMs: stable URLs, structured data, clean semantics, robots controls, and machine-readable endpoints." It calls the term "a loose umbrella term" and adds: "None of it is a single formal standard." The page is labelled "Recommended" and was updated on 8 June 2026.
Shopify uses the words without defining them. Its free audit lives at shopify.com/agentic-readiness, under the title "Product Page Audit for Agentic Commerce", and the audit page has no sentence saying what agentic readiness is. Stripe titles a guide "Agent-ready retail in 2027" and does not define "agent-ready" either. Its nearest sentence says the approaches it describes "all share a common requirement: structured data, secure transactions, and agent-compatible checkout flows."
Read side by side, Cloudflare's version is about websites in general and leaves commerce out of the score. specification.website's is also about any site. Stripe's sentence and Shopify's tool are about selling. That split explains why two tools can give the same store very different marks: they are not measuring the same thing.
02What do the definitions have in common?
Three ideas recur, in our reading. An agent must be allowed in, through robots.txt and bot rules. It must be able to read the site as data, through structured data, Markdown or other machine-readable files. And for a store, it must be able to act, which means checkout and payment. Cloudflare and specification.website both cover the first two. Shopify's audit reads structured data and robots.txt. Stripe's sentence names structured data and checkout, and Cloudflare reports commerce checks without scoring them.
How common those signals are is a separate question. Cloudflare published a snapshot from its Radar data for the top 200,000 domains in the same April post.
How rare each agent signal is, by Cloudflare’s count
Showing: robots.txt
robots.txt
Cloudflare: “78% of sites have one”
In our words: A text file of crawl rules at the root of a site.
Cloudflare’s figures for the top 200,000 domains, as published in its post of 17 April 2026 (the live Radar chart updates weekly). They are not Storewalk data.
The gap between robots.txt and everything else is wide. By Cloudflare's count, "78% of sites have one", while Markdown on request "passes on 3.9% of sites" and MCP server cards and API catalogs "together appear on fewer than 15 sites in the entire dataset". Those are Cloudflare's figures, not ours, and they describe websites in general, not only stores.
03How does Storewalk measure agent readiness?
Storewalk is ours, so everything in this section comes from our own product copy and code. We name the file each sentence comes from in the figure below.
The scan starts by asking permission. Our site says: "Before any page, we read what your robots.txt allows and keep to a polite pace." It then reads the store in four steps, in the order a shopping agent would use it:
Find: "An agent has to find your shop and your products without a person pointing the way."
Compare: "It has to read price, stock, delivery and returns as text it can trust."
Buy: "It has to reach the payment step as a guest, with nothing in the way."
After the order: "It has to track the order, start a return and reach support."
A few checks sit "beside the path": they "do not block a purchase, but decide how agents are treated and seen."
How a Storewalk score is made, stage by stage (ours)
Showing: Ask permission
Ask permission
In our words
Before it reads any page, the scan reads your robots.txt and keeps to its rules and a slow pace.
Our product copy
“Before any page, we read what your robots.txt allows and keep to a polite pace.”
packages/core/data/agent-path.json
Walk the four steps
In our words
It reads the store in the order a shopping agent would use it: find, compare, buy, and after the order. A few checks sit beside the path.
Our product copy
“The scan reads your store in the order an agent would use it: find, compare, buy, and after the order.”
apps/web/app/page.tsx
Keep the evidence
In our words
Every check records what it read on your store, or what it could not find, so a result can be traced back to a page.
Our product copy
“For every check, the page or markup it read on your shop, or what it could not find.”
apps/web/components/landing/WhatComesBack.tsx
Skip, with a reason
In our words
A check that could not run is marked skipped and left out of the score. It is never counted as a pass or a fail.
Our product copy
“When bot management answers instead of the shop, or a page never loads, the check is skipped with the reason attached. We do not guess its result or average it into the score.”
apps/web/components/landing/CannotTell.tsx
Ready, and Basics apart
In our words
Checks for things built for agents make up Ready. Ordinary web hygiene goes into a separate Basics figure and does not move Ready.
Our product copy
“Standard web hygiene. Not part of your agent-readiness score.”
If the scan could not read enough of the store, the report prints why instead of a number.
Our product copy
“When the evidence is thin, there is no score.”
apps/web/components/landing/CannotTell.tsx
Safe, for the owner
In our words
A second score, about telling real agents from fake ones, needs proof that the domain is yours first.
Our product copy
“A second score, Safe, shows whether your shop can tell a real agent from a spoofed one. You need to verify your domain first”
apps/web/components/landing/WhatComesBack.tsx
What comes back is a Ready score. Our landing page describes it as "A score out of 100 for whether an agent could find your products, read prices and stock, and check out as a guest. We read it from your own pages and show the step where each point was lost."
Ready counts only the checks for things built for agents, such as an llms.txt file, a UCP profile or machine-readable shipping terms. Ordinary web hygiene that most stores already have, such as a sitemap, product structured data or guest checkout, goes into a separate figure called Basics. The report labels it "Standard web hygiene. Not part of your agent-readiness score." We made that split on purpose. A tidy, conventional store can score well on Basics and low on Ready, and both numbers are true at once.
A second score, Safe, "shows whether your shop can tell a real agent from a spoofed one". It needs you to verify your domain first, "because it reads things we should only read for the owner."
Two rules decide when there is no number. A check that cannot run is skipped: "When bot management answers instead of the shop, or a page never loads, the check is skipped with the reason attached. We do not guess its result or average it into the score." And for the whole report: "When the evidence is thin, there is no score." The report "keeps the space where the score would be and prints the reason it was withheld." A blank with a reason is less satisfying than a number, but it has the advantage of being true.
The scanner reads at most 75 pages, at 2 requests a second, sends one user agent and deletes the data after 30 days. Its checkout checks are passive: it reads checkout pages and agent checkout endpoints, and never adds anything to a cart. You can see every check, grouped by step, in our check catalogue.
04How do other tools measure it?
Every tool we know of reports in its own unit, and the units do not convert. Pick one below to see what it returns, in its own words.
Five tools, five scales: what each agent-readiness score is made of
Showing: isitagentready (Cloudflare)
isitagentready (Cloudflare)
What comes back (our summary)
A level from 0 to 5, and a percentage
In its own words
“Returns a readiness level (0-5), the status of all 23 checks, and fix instructions”
Worth knowing
Commerce checks (x402, UCP, ACP) are reported but “do not currently count towards the score”.
Five units for one phrase. A 60 on one is not a 60 on another, so compare a store with itself over time, on one tool.
isitagentready, Cloudflare's scanner, returns "a readiness level (0-5), the status of all 23 checks, and fix instructions", and shows a percentage beside it. Lighthouse's Agentic Browsing category, from Google's Chrome team, "does not have a weighted average score from 0 to 100"; it reports how many of its checks your site passes. UCP Checker scores a store's UCP profile out of 100, weighted "30% Discovery 40% Conformance 30% Capabilities". Shopify's audit reads one product page and says "Each of the … factors below is weighted equally". Storewalk gives Ready by step, with Basics beside it.
So a 60 from one tool and a 60 from another are two different facts. The fair comparisons are a store against itself over time on one tool, or two stores on the same tool. Storewalk does the second for you if you name up to two competitors when you start a scan.
Alternatives to isitagentready describes eleven tools in detail, with what each one checks and what it says it cannot tell you.
05What should you check first?
Whatever tool you use, these are the checks we would start with. Each links to our check page, which explains what it reads.
Your platform publishes a UCP profile or ACP endpoint, which our agent checkout check looks for. UCP explained covers what that profile contains.
06Frequently asked questions
What does agent readiness mean?
How well an AI agent can use a site for a person: find it, read it as data, and, for a store, buy from it. The term has no standard definition. Cloudflare says its score shows "how well their websites support AI agents"; specification.website calls it "the set of choices that make a site legible to AI agents and LLMs".
Is agent readiness the same as SEO?
They overlap. Structured data and robots.txt matter to both. Agent readiness adds things search engines do not need, such as Markdown on request, discovery files and an agent checkout. In Storewalk's report, the overlap with ordinary web hygiene sits in Basics, apart from the Ready score.
Can I compare scores from different tools?
No. isitagentready uses levels and a percentage, Lighthouse a fraction, UCP Checker a weighted score out of 100, Shopify equal-weight factors for one page, and Storewalk a Ready score by step. Compare a store against itself on one tool.
Why does a scan skip a check?
In Storewalk, a check is skipped when it could not read what it needs, for example when bot management answers instead of the shop or a page never loads. The reason is attached, and the check is left out of the score rather than guessed.