Each check belongs to one of four categories. The categories add up to 100 points, so the score reads out of 100.
A cap of 59 keeps a site below Fair when agents are turned away or can't finish a task. A cap of 79 keeps it below Good when agents get in but can't rely on what they find. The report says when a cap is holding the score down, and which fix lifts it.
Can AI agents get in? robots.txt, bot protection, challenges and whether key pages open for them. 9 checks share its points by weight (20 in all).
Bot protection lets AI agents through
Weight 4
Worth up to 5 points
Caps the score at 59 on a fail
Usually fixed by Bot protection or CDN admin
Effort Hours
When bot protection blocks AI agents, a person asking ChatGPT or Perplexity about you gets nothing back, or gets a competitor instead. Our test requests come from our servers, not the AI companies', so we also send one claiming to be Googlebot. If that's turned away too, your protection is checking where requests really come from, which is the right setup: we can't tell whether the real agents get through, so the check is skipped and never counts against you. A request that gets no answer within 15 seconds isn't counted as a block either; if every one times out, the check is skipped, and it's worth looking at your page speed.
How to fix it →
AI agents aren't blocked or shown a CAPTCHA at the cart and checkout
Weight 3
Worth up to 3.8 points
Caps the score at 59 on a fail
Usually fixed by Bot protection or CDN admin
Effort Hours
A CAPTCHA or block at cart or checkout stops an agent at the last step, after it has done all the work, so the order is lost. If your bot protection checks where requests really come from, our test requests can't show whether real assistants get through, so a block isn't counted.
How to fix it →
AI agents can open your product, pricing and cart pages
Weight 3
Worth up to 3.8 points
Caps the score at 59 on a fail
Usually fixed by Bot protection or CDN admin
Effort Hours
Product, pricing and cart pages are what an assistant needs to answer “how much is it?” or help someone buy. If agents can't open them, the sale goes elsewhere. Each page is checked against the robots.txt of the address it's on, so www and non-www rules both count. If your bot protection checks where requests really come from, our test requests can't show whether real assistants get through, so a block isn't counted. A page that takes more than 15 seconds to answer our assistant request is left untested rather than counted as blocked.
How to fix it →
robots.txt lets AI assistants and search agents in
Weight 3
Worth up to 3.8 points
Usually fixed by SEO or content team
Effort Hours
If robots.txt shuts out AI assistants and AI search, they can't read your pages, so they leave you out of answers and recommendations.
How to fix it →
No CAPTCHA or challenge on arrival
Weight 2
Worth up to 2.5 points
Caps the score at 59 on a fail
Usually fixed by Bot protection or CDN admin
Effort Hours
Agents can't solve CAPTCHAs. A challenge on arrival stops them before they see anything.
How to fix it →
Shoppers can check out without an account
Weight 2
Worth up to 2.5 points
Caps the score at 79 on a fail
Usually fixed by E-commerce platform admin
Effort Days
An agent buying for someone can't easily create an account for them. Guest checkout lets the purchase go through. Most checkouts only open with something in the cart, so AgentScore adds one item, once, with the add-to-cart button (never “Buy now”), to look; it never fills in the form or orders anything.
How to fix it →
Agents can check out through an agentic commerce protocol
Weight 1
Worth up to 1.3 points
Usually fixed by E-commerce platform admin
Effort Weeks
Agentic commerce protocols let assistants like ChatGPT complete a purchase safely on a shopper's behalf. Early support puts you in front of those shoppers.
How to fix it →
Agents can use your site through MCP or WebMCP
Weight 1
Worth up to 1.3 points
Usually fixed by Developer
Effort Weeks
MCP or WebMCP tools let agents search, check stock and build a cart directly, which is faster and more reliable than clicking through pages. A server description over 1 MB still counts: we read its start.
How to fix it →
llms.txt guide for AI
Weight 1
Worth up to 1.3 points
Usually fixed by SEO or content team
Effort Hours
An llms.txt file gives AI a quick, accurate summary of your business and key pages. It's new, so few competitors have one yet.
How to fix it →
Can they read what they find? Content without JavaScript, structured data, prices and product facts. 8 checks share its points by weight (20 in all).
Content loads without JavaScript
Weight 4
Worth up to 5 points
Caps the score at 79 on a fail
Usually fixed by Developer
Effort Weeks
Most AI crawlers and many assistants don't run JavaScript. If your content only appears after scripts run, they see an empty page.
How to fix it →
Price, stock and name agree everywhere agents read them
Weight 3
Worth up to 3.8 points
Caps the score at 79 on a fail
Usually fixed by Developer
Effort Days
Agents read the price from your product data, the page and its sharing tags. When they disagree, assistants quote the wrong price or say something is in stock when it isn't, and shoppers who arrive expecting it leave.
How to fix it →
Product pages give price and stock in a form agents can read
Weight 3
Worth up to 3.8 points
Usually fixed by Developer
Effort Days
If price and stock aren't machine-readable on product pages, assistants can't say whether you have the item or what it costs.
How to fix it →
Structured data describes your business and products
Weight 3
Worth up to 3.8 points
Usually fixed by Developer
Effort Days
Structured data tells agents exactly what you sell, at what price, and whether it's in stock, so they quote you correctly.
How to fix it →
Clear page title, description, and headings
Weight 2
Worth up to 2.5 points
Usually fixed by SEO or content team
Effort Hours
Clear titles, descriptions and headings are how agents work out what a page is about and whether it answers the question they were asked.
How to fix it →
Prices are in the page HTML
Weight 2
Worth up to 2.5 points
Usually fixed by Developer
Effort Days
When prices load with JavaScript, agents that don't run it see a product with no price, and often skip it. Prices are recognised in the currencies stores use most, with the symbol before or after the amount.
How to fix it →
Valid sitemap
Weight 2
Worth up to 2.5 points
Usually fixed by SEO or content team
Effort Hours
A sitemap helps crawlers and agents find all your important pages, including ones that aren't linked from the home page. A sitemap over 50 MB or 50,000 URLs, the most crawlers and agents will read, is flagged: split it into smaller sitemaps so nothing gets missed.
How to fix it →
Images have text descriptions
Weight 1
Worth up to 1.3 points
Usually fixed by SEO or content team
Effort Days
Agents can't see images. Alt text tells them what a product photo shows.
How to fix it →
Can they find their way and press things? Named buttons, labelled fields, pop-ups and links to key pages. 8 checks share its points by weight (22 in all).
Buttons and links have names
Weight 4
Worth up to 3.6 points
Usually fixed by Developer
Effort Days
Agents choose what to click by its name. A button with no text or label is invisible to them.
How to fix it →
Pop-ups and banners can be dismissed by agents
Weight 4
Worth up to 3.6 points
Usually fixed by Developer
Effort Days
A pop-up or cookie banner an agent can't close blocks everything behind it, so the journey ends there.
How to fix it →
Agents can find and press your add-to-cart or sign-up button
Weight 3
Worth up to 2.7 points
Caps the score at 79 on a fail
Usually fixed by Developer
Effort Days
If an agent can't find or press “Add to cart”, “Sign up” or “Book now”, it can't convert, however good the rest of the page is. If the product we check is sold out, we try up to two others, and if they're all sold out this check is skipped, so it never counts against you.
How to fix it →
Cart and checkout fields are labelled for agents
Weight 3
Worth up to 2.7 points
Usually fixed by Developer
Effort Days
Checkout is where agents most often fail. Clearly labelled fields let them enter details correctly and finish the order. When the empty cart shows no form, AgentScore adds one item to the cart, once, with the add-to-cart button (never “Buy now”), to see the checkout; it never fills in the form or orders anything.
How to fix it →
Clickable things are real buttons and links
Weight 2
Worth up to 1.8 points
Usually fixed by Developer
Effort Days
Agents look for real buttons and links. Clickable areas built from plain boxes often aren't recognized as clickable at all.
How to fix it →
Form fields are labelled
Weight 2
Worth up to 1.8 points
Usually fixed by Developer
Effort Days
Agents need labels to know what goes in each field. Unlabelled forms get filled in wrongly or abandoned.
How to fix it →
Key pages are linked from the home page
Weight 2
Worth up to 1.8 points
Usually fixed by Developer
Effort Hours
Agents start at the home page and follow links. If products, pricing and cart aren't linked there, agents may never find them.
How to fix it →
Page has main and navigation landmarks
Weight 2
Worth up to 1.8 points
Usually fixed by Developer
Effort Hours
Main and navigation landmarks help agents skip menus and go straight to the content.
How to fix it →
Can an AI agent actually finish a job? The free scan has a real agent try a typical task, like adding a product to the cart. One check carries all of its points.
When we change a check, a weight or a cap, the scoring version changes, and every report shows the version it was scored with. Scores from different versions aren't directly comparable, so rescan to compare like with like. The changelog and blog explain what changed and why.
