Shopify Product Page Audit for Agentic Commerce
Shopify runs this one itself, at shopify.com/agentic-readiness, so a Shopify merchant is likely to meet it before any other scanner. It has a head start: its maker already hosts their store. The page says: "We check your product page for the structured data that AI agents read to answer shoppers' questions." It takes one product page, not a whole store: "Enter a full product page URL".
The page's result strings name 11 factors. One is whether robots.txt lets in the 10 search and AI user agents it checks. Nine are schema.org fields: the product's name, description, price and currency, availability, variants, images and per-variant images, its aggregate rating, and the shop's Organization markup. The last is the page title, which should run 30 to 60 characters. Each factor is marked "AI ready", "Needs improvement", "Fail" or "Not detected", and all 11 are "weighted equally" into an "Agent discoverability score". The page adds that these are "informational signals" that "do not guarantee your products will be surfaced by AI agents."
None of the 11 factors is llms.txt, an MCP server, a UCP profile, the cart or checkout. The matrix below marks those "not stated", not "no", because we read the factor list from the page's code and could not run a scan ourselves. The form has no login or email field, and we saw no paywall. The page does not say whether it accepts a product page from a store that is not on Shopify. It suits a merchant who wants a fast check that one product page carries the data an agent reads, and that robots.txt lets the main AI crawlers in.
Fudge's Shopify AI Readiness Checker
Fudge IT, Inc. makes Fudge, an AI page builder and theme editor for Shopify ("a Shopify-native AI app used by hundreds of stores"), and offers the checker on its site: "Completely free. No credit card required." It covers Shopify stores only ("Not right now. We focus on Shopify-specific signals and best practices."), reads public pages only, and asks for your email before it sends the report, "to protect our systems from spam requests".
Its page lists four groups of checks. The basics are robots.txt and bot permissions, "Shopify MCP discoverability", mobile speed, "llms.txt presence and correctness" and "Shopify Catalog eligibility signals". Shopify Catalog is how Shopify passes products on, in its words "to distribute to AI agents". Then come policies, About, contact and FAQ pages. On product pages it reads "AI crawlability and rendered content", GTIN, UPC or EAN codes, specs, review structured data, shipping and returns, and alt text. Last come use-case, comparison and collection pages. You get an "AI Readiness Score" and a list of fixes in priority order in about 60 seconds, and the page notes, "We sample products for speed." The page names no user agent for its crawler. It suits a Shopify merchant who wants a wider sweep than one product page and is happy to give an email address for it.
The Shopify App Store also has newer apps that score a store's readiness for agents: Agentic Commerce Readiness, AgentSight, StoreHQ, Sichta and a8s Agent Ready. Each installs into a Shopify store. They launched between May and July 2026, and none had a review when we checked, so we leave them out of the comparison for now.
Lighthouse Agentic Browsing
Lighthouse is a page audit tool whose source sits in the GoogleChrome repository on GitHub, under Apache-2.0; its latest release, v13.6.0, came out on 27 October 2026. Chrome for Developers says: "The Agentic Browsing category and WebMCP support are experimental and based on proposed standards. Testing this category requires Chrome 150 or later, and WebMCP audits require registering for the WebMCP origin trial." (An origin trial, in our words, is Chrome's sign-up for trying an unreleased feature on a site.)
In v13.6.0 the category's audits are the agent accessibility tree, three WebMCP audits, cumulative layout shift (how far the page moves while it loads, in our words), llms.txt, and ard-schema, which is in the source but not on the docs page. The accessibility tree is the structure a browser builds for assistive technology, covered in how an AI agent reads a web page. The result is "A ratio showing how many agentic readiness checks your site passes", not "a weighted average score from 0 to 100." Its audit list has nothing on products, checkout or payment, so it is a tool for a developer looking at how one page behaves for an agent in the browser.
Google Rich Results Test
The Rich Results Test checks structured data (a machine-readable description of a page, in JSON-LD, RDFa or Microdata) against the rich result types Google supports, among them "Merchant listings", "Product snippet", "Return policy" and "Shipping policies". It "uses the rendered source code" and fetches as Google-InspectionTool, so "it can be blocked by a robots.txt file".
It is not an agent-readiness tool. We include it because product structured data is where the scanners differ: isitagentready does not check it, and Shopify's audit, Fudge's checker, Storewalk, StartupHub.ai and AgentReady say they do. Use it to see whether Google reads your product, shipping and return markup.
UCP Checker
UCP Checker's terms say it "is a trading name of PactMode Ltd" and "is not affiliated with, endorsed by, or operated on behalf of any merchant, platform, or the maintainers of the UCP specification." A UCP profile is the file at /.well-known/ucp where a store describes what it supports; UCP explained goes through it.
Its about page says: "We read each store's /.well-known/ucp and check it against the release it declares: services, capabilities, payment handlers and keys." It also reads robots.txt and calls the services a profile declares. Its UCP Score runs from 0 to 100, weighted "30% Discovery 40% Conformance 30% Capabilities", and Discovery takes in llms.txt, the sitemap and three other files outside UCP. Its crawler, "UCPCheckerBot/1.0", signs with Web Bot Auth, does not fetch the profile if robots.txt disallows it, allows 2 seconds for the profile request ("about the time an agent waits"), and "never carts or checks out." The check is free and needs no account. It suits a store whose platform publishes a UCP profile.
StartupHub.ai Agent Readiness scanner
Its page reads: "Paste a website, REST API or MCP endpoint and get a 0 to 100 score for AI agent readiness, a letter grade, and a fix prompt for every failed check." It gives "about 40 checks" in one place and "44 checks" in another, in six categories, commerce among them. The commerce checks look for "x402, MPP, AP2 or ACP discovery, x402-mesh, Offer or Product JSON-LD with a price, a payment link reachable from the homepage and whether /checkout or /pricing exists."
You can "Scan competitors or peers alongside your site and compare scores", and a Live Agent Run "starts six agent frameworks on StartupHub.ai hardware". The scanner "is free and needs no signup". We have not verified which company runs it.
AgentReady
AgentReady is a command-line tool published by swarmclawai on GitHub under Apache-2.0. Its version, 0.1.0-alpha.0, is an alpha that the README calls "an early CLI release". It reads robots.txt, the sitemap, llms.txt, llms-full.txt and Markdown fallbacks; "structured data, product schema, service schema, price, inventory, cart, checkout, shipping, tax, refund, cancellation, and support signals"; A2A agent cards; and "HTTP 402 and x402 metadata hints". It "does not submit forms, make purchases, trigger paid actions", and prints a score and findings as text, JSON, Markdown or HTML. Developers who want a scanner they can run locally and read, commerce signals included, are its audience.
agent-readiness-mcp
Forgemesh Labs publishes "the free, open-source Agent Readiness MCP" under MIT, labelled "Beta. Experimental ASO scanner". It is an MCP server, so in our words an AI assistant can call it as a tool. It reads "34 signals across 6 pillars", Commerce among them ("x402, MPP, UCP, ACP, machine-readable pricing"), and returns what it calls an "ASO Score", from 0 to 100, with a level; "the auto-verifiable maximum is 89/100." It fits someone who already works inside an AI assistant.
UCP conformance suite and ucp-schema
Both come from the UCP authors' GitHub organisation, under Apache-2.0. The conformance suite holds "language-agnostic integration tests that run against a running UCP Merchant Server instance", which "must be populated with test data"; its tests cover checkout, orders, discounts, fulfilment, AP2, card credentials and webhooks. ucp-schema is a "CLI and library for Universal Commerce Protocol (UCP) schemas" that can validate ("check payload against schema"). Neither scans a live public store; they are for whoever builds or runs a UCP server.
Storewalk (ours)
Storewalk is ours. Its site says "The scan reads your store in the order an agent would use it: find, compare, buy, and after the order", and its check catalogue is grouped by those steps. Its checks run from robots.txt, llms.txt, product structured data and content without JavaScript to discovery files and WebMCP tools. For buying it looks for guest checkout, no CAPTCHA before payment, agent payment tokens at the payment provider, and an agent checkout endpoint, which it looks for in /.well-known/ucp and ACP's proposed /.well-known/acp.json. After the order it reads one thing: support contact details in the markup.
The checkout checks are passive. They read checkout pages and agent checkout endpoints, and the scanner does not add anything to a cart, so its cart is always empty. When the empty cart sends it away from checkout, the guest-checkout check is skipped with a reason.
The report gives a Ready score out of 100 and shows "the step where each point was lost". You can name up to two competitors and see their Ready score beside yours. The scanner reads at most 75 pages at 2 requests a second, counts a page opened in a real browser against that budget, reads robots.txt first, sends one user agent, "never disguised", and deletes the data after 30 days. When bot management answers instead of the shop, the check is skipped with the reason attached, and "When the evidence is thin, there is no score." You need no account to scan, but the full report opens after you give an email address and a six-digit code, and the second score, Safe, needs an account in which you verify your domain. It is not open source, and the report offers a free consultation with us. We built it for store owners who want to see, step by step from finding a product to paying, what an agent would find on their store and where the signals run out.