{
 "version": "1",
 "collection": "glossary",
 "about": "Agent readiness terms in plain English.",
 "updated": "2026-10-09",
 "count": 42,
 "items": [
  {
   "id": "ai-agent",
   "url": "https://ghostagentlab.com/glossary/ai-agent/",
   "name": "AI agent",
   "definition": "Software that uses an AI model to do something on the web for a person or a company: read a page, answer a question, compare products or complete a task. In our guides, “agent” covers everything from crawlers to agents that shop.",
   "why": "AI agents are becoming a new way customers find, compare and buy. When someone asks an assistant for the best running shoe under $150, the answer depends on which sites its agents could read and use. A site agents can't use doesn't get a complaint or a support ticket; it simply isn't recommended. Knowing which agents visit, and whether they succeed, is the starting point.",
   "example": "A shopper asks an assistant to compare three jackets, and within seconds its agent opens a category page, three product pages and the returns policy on each store it is considering.",
   "howToCheck": "Run a free AgentScore scan at /agentscore/ to see whether AI agents can get in, read your pages and find their way around. The free AI crawler checklist at /tools/ai-crawler-checklist/ lists the robots.txt names the major AI companies use, sorted by kind of agent.",
   "related": [
    "agent-traffic",
    "robots-txt",
    "agent-readiness",
    "journey"
   ],
   "section": "The agents",
   "guide": null,
   "modified": "2026-10-09"
  },
  {
   "id": "training-crawler",
   "url": "https://ghostagentlab.com/glossary/training-crawler/",
   "name": "Training crawler",
   "definition": "A bot that collects pages to train AI models, such as GPTBot or ClaudeBot. Blocking it doesn't stop AI assistants visiting for a person.",
   "why": "Training crawlers are the agents most sites argue about, because they take content without sending visitors back. Blocking them is a reasonable business choice, but it's often done with rules broad enough to block the assistants that do bring customers. Decide on training crawlers separately, by name, so a choice about AI training doesn't quietly cost you AI search and shopping traffic.",
   "example": "A publisher adds Google-Extended and Applebot-Extended to robots.txt to opt out of AI training. These are control names rather than separate crawlers, so Googlebot and Applebot keep crawling for normal search.",
   "howToCheck": "Paste or fetch your robots.txt into the robots.txt tester at /tools/robots-txt/: it lists each AI training crawler, whether it is allowed or blocked, and the rule that decides it. AgentScore also reports blocked training crawlers, but doesn't take points off for them.",
   "related": [
    "robots-txt",
    "ai-search-crawler",
    "crawls-per-ai-referral",
    "verification"
   ],
   "section": "The agents",
   "guide": null,
   "modified": "2026-10-09"
  },
  {
   "id": "ai-search-crawler",
   "url": "https://ghostagentlab.com/glossary/ai-search-crawler/",
   "name": "AI search crawler",
   "definition": "A bot that indexes pages so AI search and assistants can cite and link to them, such as OAI-SearchBot or PerplexityBot.",
   "why": "AI search crawlers decide whether your pages can be cited and linked in AI answers, which is where a growing share of product research now starts. If they're blocked, or can't read your pages, assistants recommend competitors they could read instead. Most sites should let them in, and check now and then that robots.txt and bot protection still do.",
   "example": "A store copies a long \"block all AI\" list into robots.txt to stop training, but the list also names OAI-SearchBot and PerplexityBot, so the store drops out of AI search answers as well.",
   "howToCheck": "The robots.txt tester at /tools/robots-txt/ shows whether each AI search crawler is allowed, and the What does ChatGPT see? tool at /tools/ai-view/ can fetch any page as OAI-SearchBot to show what it gets. Ask your CDN admin to confirm AI search is in an allowed bot category.",
   "related": [
    "robots-txt",
    "raw-html",
    "bot-protection",
    "xml-sitemap"
   ],
   "section": "The agents",
   "guide": null,
   "modified": "2026-10-09"
  },
  {
   "id": "assistant-fetcher",
   "url": "https://ghostagentlab.com/glossary/assistant-fetcher/",
   "name": "Assistant fetcher",
   "definition": "An AI assistant fetching a page in real time because a person asked it something, such as ChatGPT-User or Claude-User. Each visit stands in for a customer.",
   "why": "An assistant fetcher is the closest thing to a customer in your agent traffic: each visit is a person asking a question that your page could answer. If bot protection turns it away or the page needs JavaScript it doesn't run, that person gets an answer built from someone else's site. These are the agent visits to protect first.",
   "example": "A shopper asks ChatGPT \"what's the return window at this store?\" and ChatGPT-User fetches the returns page there and then. If bot protection answers with a 403, the assistant tells the shopper it couldn't reach the site.",
   "howToCheck": "Enter a product or policy page into What does ChatGPT see? at /tools/ai-view/ and view it as ChatGPT-User, Claude-User or Perplexity-User, next to what your visitors see. Your CDN or server logs show whether these agents get 403, 429 or challenge responses.",
   "related": [
    "raw-html",
    "bot-protection",
    "ai-referral",
    "rate-limit"
   ],
   "section": "The agents",
   "guide": null,
   "modified": "2026-10-09"
  },
  {
   "id": "browser-agent",
   "url": "https://ghostagentlab.com/glossary/browser-agent/",
   "name": "Browser agent",
   "definition": "An agent that opens your site in a real browser and clicks, types and scrolls like a person, to book, sign up or buy. Also called a computer-use agent or autonomous agent.",
   "why": "Browser agents are the ones that can actually book, sign up or buy for a customer, so they're where agent traffic turns into revenue. They use your site the way a person does, which means pop-ups, unlabelled buttons and fiddly forms stop them just as they frustrate shoppers. The only way to know whether they can finish is to send one through your key journeys.",
   "example": "On a fictional running store, a browser agent finds the shoe and price, but the sizes are plain boxes rather than buttons and a cookie banner covers \"Add to cart\", so it may never pick a size or add the shoe.",
   "howToCheck": "Give an AI assistant's agent mode your most valuable journey, such as \"find a coffee maker and add it to the cart, don't buy anything\", and watch where it hesitates. A free AgentScore scan checks the things browser agents trip on: unnamed buttons, pop-ups and CAPTCHAs.",
   "related": [
    "accessibility-tree",
    "overlay",
    "guest-checkout",
    "journey"
   ],
   "section": "The agents",
   "guide": null,
   "modified": "2026-10-09"
  },
  {
   "id": "user-agent",
   "url": "https://ghostagentlab.com/glossary/user-agent/",
   "name": "User agent",
   "definition": "The name a visitor sends with every request to say what it is, such as a browser version or GPTBot. Easy to fake, so it shouldn't be trusted on its own.",
   "why": "User agents are how most analytics, logs and bot rules decide what a visitor is, so it matters that they can be faked. A rule that trusts the name alone lets scrapers in by pretending to be Googlebot, while a rule that blocks by name can turn away real assistants. Pair names with verification before you let a visitor in or count it as a real agent.",
   "example": "A server log line shows a request for a product page with the user agent \"compatible; ChatGPT-User/1.0\" and a 403 status: the name says who it claims to be, and the status says it was blocked.",
   "howToCheck": "Ask whoever runs your CDN or servers for a sample of log lines and look at the user agent field. To test whether one of them is genuine, paste its IP address and user agent into Is this bot real? at /tools/verify-bot/.",
   "related": [
    "verification",
    "impostor",
    "agent-traffic",
    "web-bot-auth"
   ],
   "section": "The agents",
   "guide": null,
   "modified": "2026-10-09"
  },
  {
   "id": "robots-txt",
   "url": "https://ghostagentlab.com/glossary/robots-txt/",
   "name": "robots.txt",
   "definition": "A file at the root of your site that tells crawlers and agents which pages they may visit. Well-behaved agents follow it; scrapers ignore it. See robots.txt for AI agents.",
   "why": "robots.txt is usually the first thing an agent reads, so one line in it can decide whether AI assistants can visit your site at all. In our study of 255 online stores, robots.txt blocked AI assistants or AI search on only 4%, but it's an easy place for an old rule to block more than anyone intended. Review it with your SEO or content owner whenever your AI policy changes.",
   "example": "A store adds User-agent: GPTBot followed by Disallow: / to keep OpenAI's training crawler off the whole site. ChatGPT-User and OAI-SearchBot have no group of their own, so they still follow the * group and can keep reading the store for customers.",
   "howToCheck": "Open yourdomain.com/robots.txt and look for a leftover Disallow: / that blocks everything. The robots.txt tester at /tools/robots-txt/ fetches your file and shows which AI agents each page allows or blocks, and you can change the page to /products/ or /cart. AgentScore's robots.txt check fails when three or more AI assistants and search agents are blocked from your home page.",
   "related": [
    "training-crawler",
    "ai-search-crawler",
    "bot-protection",
    "xml-sitemap"
   ],
   "section": "Getting in: access",
   "guide": "https://ghostagentlab.com/articles/robots-txt-ai-agents/",
   "modified": "2026-10-09"
  },
  {
   "id": "bot-protection",
   "url": "https://ghostagentlab.com/glossary/bot-protection/",
   "name": "Bot protection",
   "definition": "Services such as Cloudflare, Akamai or DataDome that block or challenge automated visitors. Set too strictly, they turn away AI assistants as well as scrapers. See bot protection and CAPTCHAs.",
   "why": "Bot protection is the most common reason AI assistants are turned away from sites that want them. In our study of 255 online stores, 13% blocked or challenged AI assistants somewhere. The fix is usually a settings change, not a rebuild: allow the agents you want, verified, and keep blocking the rest. It's normally a job for your security or CDN team.",
   "example": "A store's bot protection doesn't block agents outright, but serves them a product page that only says \"Please enable JavaScript to continue\", so an agent learns nothing about the product, price or stock.",
   "howToCheck": "A free AgentScore scan requests your home page as a browser and as ChatGPT-User, Claude-User, Perplexity-User and OAI-SearchBot, and flags an agent that is blocked, challenged or gets under half the content. You can see a single page the same way with What does ChatGPT see? at /tools/ai-view/. Ask your CDN admin which AI categories are allowed and what any \"block AI\" setting covers.",
   "related": [
    "cdn",
    "captcha-or-challenge",
    "verification",
    "assistant-fetcher"
   ],
   "section": "Getting in: access",
   "guide": "https://ghostagentlab.com/articles/bot-protection-ai-agents/",
   "modified": "2026-10-09"
  },
  {
   "id": "cdn",
   "url": "https://ghostagentlab.com/glossary/cdn/",
   "name": "CDN",
   "definition": "Content delivery network: the service that sits in front of your site and delivers it quickly around the world. It's often where bot rules live, and where agent traffic is recorded. See setting up your CDN for AI agents.",
   "why": "Your CDN is where many agent decisions are made and recorded: it often holds the bot rules, and its logs are the most complete record of which agents visit. That makes it the place to fix blocking problems, and the place to measure agent traffic your analytics can't see. Ask whoever runs it which bot settings are switched on.",
   "example": "A CDN has a broad \"block AI bots\" rule sitting above the rule that allows verified AI assistants, so the allow rule never runs and ChatGPT-User is turned away.",
   "howToCheck": "Ask your CDN admin for the list of bot rules in the order they run, and whether AI assistants and AI search are in an allowed verified bot category. Ask too whether the logs record the action taken and the rule that took it, and whether new rules run in log-only mode before they block.",
   "related": [
    "bot-protection",
    "agent-traffic",
    "rate-limit",
    "web-bot-auth"
   ],
   "section": "Getting in: access",
   "guide": "https://ghostagentlab.com/articles/cdn-settings-ai-agents/",
   "modified": "2026-10-09"
  },
  {
   "id": "captcha-or-challenge",
   "url": "https://ghostagentlab.com/glossary/captcha-or-challenge/",
   "name": "CAPTCHA or challenge",
   "definition": "A test that tries to tell people from bots, such as picking out images or waiting on a “checking your browser” page. Agents can't pass them, so one on arrival or at checkout ends the visit.",
   "why": "A CAPTCHA ends an agent's visit on the spot, because agents can't solve them. In our study of 255 online stores, 6% showed one on the home page, and 15% of the carts and checkouts we could load showed one to every visitor. If you need one, show it only when something looks wrong, not to everyone at checkout.",
   "example": "A store shows a puzzle CAPTCHA to every visitor at checkout, so an agent that has chosen a product and filled the cart stops at the last step. A risk-based challenge that escalates only when traffic looks abusive would let it through.",
   "howToCheck": "Open your home page, a product page and the checkout in a private window and note any challenge or \"checking your browser\" page. A free AgentScore scan checks for a CAPTCHA or challenge on arrival, and whether agents are blocked or shown one at the cart and checkout.",
   "related": [
    "bot-protection",
    "guest-checkout",
    "browser-agent",
    "task-completion"
   ],
   "section": "Getting in: access",
   "guide": null,
   "modified": "2026-10-09"
  },
  {
   "id": "rate-limit",
   "url": "https://ghostagentlab.com/glossary/rate-limit/",
   "name": "Rate limit",
   "definition": "A cap on how many requests a visitor can make in a given time. Fair limits stop scrapers without turning away agents. See rate limits for AI agents.",
   "why": "Rate limits protect your site from being overloaded, but a limit set too tight turns away agents that are doing legitimate research for customers. A fair limit tells an agent when to come back instead of refusing it outright, and gives verified agents room to work. Your developers or CDN team can usually adjust this in an afternoon.",
   "example": "An assistant comparing three jackets fetches a category page, three products and a returns policy in seconds. A per-IP limit answers with 403 Forbidden and the shopper hears the store is down, where 429 with Retry-After: 30 would have told it to wait and try again.",
   "howToCheck": "Ask for your logs by agent and look at the share of assistant requests getting 429, 403 or challenge responses: repeated ones mean questions going unanswered. A free AgentScore scan counts any error an agent gets while a browser gets through, including 429, as blocked.",
   "related": [
    "bot-protection",
    "verification",
    "assistant-fetcher",
    "agent-traffic"
   ],
   "section": "Getting in: access",
   "guide": "https://ghostagentlab.com/articles/rate-limits-ai-agents/",
   "modified": "2026-10-09"
  },
  {
   "id": "verification",
   "url": "https://ghostagentlab.com/glossary/verification/",
   "name": "Verification",
   "definition": "Checking that a visitor claiming to be a known agent really is, using the proof its operator publishes: IP ranges, reverse DNS or a signature. See how to tell if an AI crawler is real.",
   "why": "Verification lets you treat agents by who they really are, not who they say they are. Without it you have to choose between letting impostors in and turning real assistants away. With it you can welcome the agents you want and block the rest, and trust that your agent traffic numbers describe real agents.",
   "example": "A request claims to be ChatGPT-User and comes from an address on OpenAI's published GPTBot list. That doesn't verify it, because each OpenAI agent has its own list, and the address has to match the list for the agent it claims to be.",
   "howToCheck": "Paste an IP address and user agent from your logs into Is this bot real? at /tools/verify-bot/. It checks published IP lists and reverse DNS, and says when an operator publishes no proof. Ask your CDN admin whether your bot rules use verified bot categories rather than names.",
   "related": [
    "forward-confirmed-reverse-dns",
    "web-bot-auth",
    "impostor",
    "agent-traffic"
   ],
   "section": "Getting in: access",
   "guide": "https://ghostagentlab.com/articles/verify-ai-crawlers/",
   "modified": "2026-10-09"
  },
  {
   "id": "impostor",
   "url": "https://ghostagentlab.com/glossary/impostor/",
   "name": "Impostor (spoofed agent)",
   "definition": "A request that uses a trusted agent's name but fails verification. Usually a scraper trying to get past bot protection.",
   "why": "Impostors are usually scrapers that copy your prices and content, and they hide behind the names of agents you trust. Left unchecked, they inflate your agent numbers and can lead teams to block the real agent whose name they borrowed. Verifying requests separates them out, so you can block impostors without hurting genuine assistants.",
   "example": "A scraper sends GPTBot as its user agent from an address that isn't on OpenAI's published list. If your firewall lets GPTBot through by name, it gets the same pass to copy your prices.",
   "howToCheck": "Paste a suspicious IP address and user agent into Is this bot real? at /tools/verify-bot/, which has a fake GPTBot example to try. Ask your CDN admin to search your firewall rules for any that allow access by user agent alone.",
   "related": [
    "verification",
    "user-agent",
    "bot-protection",
    "agent-traffic"
   ],
   "section": "Getting in: access",
   "guide": null,
   "modified": "2026-10-09"
  },
  {
   "id": "forward-confirmed-reverse-dns",
   "url": "https://ghostagentlab.com/glossary/forward-confirmed-reverse-dns/",
   "name": "Forward-confirmed reverse DNS",
   "definition": "A verification check that an IP address's hostname belongs to the operator, and that the hostname points back to the same address.",
   "why": "This check is the reliable way to confirm a visitor really comes from the company it names, for operators that publish their hostnames. You rarely need to run it by hand: good bot protection and agent monitoring do it for you. What matters is that your rules rely on it, rather than on a name anyone can type.",
   "example": "A request says it's Googlebot from 66.249.66.1. That address looks up to crawl-66-249-66-1.googlebot.com, and that name looks up back to 66.249.66.1, so the request really is from Google.",
   "howToCheck": "Is this bot real? at /tools/verify-bot/ runs this for you: paste the IP address and user agent, and it looks up the reverse DNS name and confirms it points back to the same address. Ask your bot protection provider whether its verified bot list uses this method for operators like Google, Microsoft and Apple.",
   "related": [
    "verification",
    "impostor",
    "user-agent",
    "cdn"
   ],
   "section": "Getting in: access",
   "guide": null,
   "modified": "2026-10-09"
  },
  {
   "id": "web-bot-auth",
   "url": "https://ghostagentlab.com/glossary/web-bot-auth/",
   "name": "Web Bot Auth",
   "definition": "A new standard in which agents sign their requests cryptographically, so sites can prove who sent them. Our own scanner and Ghost Agents sign theirs. See why we sign every request.",
   "why": "Web Bot Auth could make telling real agents from impostors much simpler: a signed request proves who sent it, without keeping lists of IP addresses up to date. It's new, so support is still growing among agents and bot protection services. Ask your bot protection provider whether they check signatures, and plan to accept signed agents as they arrive.",
   "example": "Every Ghost Agent request carries a Signature-Agent header pointing to app.ghostagentlab.com, where the public key is published. A CDN fetches the key once and can confirm the request is genuine from any IP address.",
   "howToCheck": "Ask your bot protection or CDN provider whether it can verify Web Bot Auth signatures, and if so whether it's turned on. If it can't yet, nothing breaks: signed requests are ordinary requests with extra headers.",
   "related": [
    "verification",
    "browser-agent",
    "cdn",
    "ghost-agent"
   ],
   "section": "Getting in: access",
   "guide": "https://ghostagentlab.com/blog/signed-requests/",
   "modified": "2026-10-09"
  },
  {
   "id": "llms-txt",
   "url": "https://ghostagentlab.com/glossary/llms-txt/",
   "name": "llms.txt",
   "definition": "A short Markdown file at /llms.txt that tells AI what your business does and links to your key pages. An emerging standard. See how to write an llms.txt file.",
   "why": "An llms.txt file is a cheap way to tell AI assistants, in your own words, what your business does and where your most important pages are. It's an emerging standard, but adoption is moving quickly: in our study of 255 online stores, 49% already published one. It takes an hour or two to write, and your content team can own it.",
   "example": "A coffee store's llms.txt opens with its name and a one-line summary, then links its Returns page with the note \"30-day returns on unopened items and equipment\", so an agent asked about returns goes straight there.",
   "howToCheck": "Open yoursite.com/llms.txt in a browser and make sure you see plain text, not your home page or a \"not found\" page. The llms.txt generator at /tools/llms-txt/ tells you if your site already has one, and drafts one from your home page if not.",
   "related": [
    "xml-sitemap",
    "mcp",
    "structured-data",
    "robots-txt"
   ],
   "section": "Getting in: access",
   "guide": "https://ghostagentlab.com/articles/llms-txt/",
   "modified": "2026-10-09"
  },
  {
   "id": "guest-checkout",
   "url": "https://ghostagentlab.com/glossary/guest-checkout/",
   "name": "Guest checkout",
   "definition": "Letting shoppers buy without creating an account. An agent buying for someone can't easily sign up for them. See guest checkout.",
   "why": "An agent buying for a customer usually can't create an account for them, so a checkout that insists on one is where the sale ends. Guest checkout also helps people who don't want another password. If you require accounts for a reason, make sure the guest route is clear and easy to find.",
   "example": "A store's checkout offers \"Continue as guest\", but the cart page above it reads \"Sign in to check out\", so an agent stops at the cart and the working guest route behind it is never used.",
   "howToCheck": "In a private window, add an item and go to checkout: look for a guest option in words and check there's no password field on the way. A free AgentScore scan runs the \"Shoppers can check out without an account\" check, and your platform admin can tell you whether the checkout setting requires accounts.",
   "related": [
    "captcha-or-challenge",
    "journey",
    "agentic-commerce",
    "browser-agent"
   ],
   "section": "Getting in: access",
   "guide": "https://ghostagentlab.com/articles/guest-checkout-ai-agents/",
   "modified": "2026-10-09"
  },
  {
   "id": "raw-html",
   "url": "https://ghostagentlab.com/glossary/raw-html/",
   "name": "Raw HTML",
   "definition": "The page exactly as your server sends it, before any JavaScript runs. It's all that many crawlers and assistant fetchers ever see. To look at it, use your browser's “View page source”, not the inspector.",
   "why": "Raw HTML is what most AI crawlers and assistant fetchers actually read, so anything missing from it is missing from their view of your business. In our study of 255 online stores, 15% didn't have prices in the product page's HTML at all. Ask your developers to check that names, prices and key details are there before any JavaScript runs.",
   "example": "A fictional store's product page looks complete in a browser, but its raw HTML holds only an empty app container saying \"Loading...\" and a title with the store's name: no product name, no price and no sizes.",
   "howToCheck": "Open a product page, choose \"View page source\" and search for the price and product name. What does ChatGPT see? at /tools/ai-view/ shows the raw HTML an assistant gets next to your live page, so you can see what's missing.",
   "related": [
    "server-side-rendering-and-static-generation",
    "assistant-fetcher",
    "structured-data",
    "accessibility-tree"
   ],
   "section": "Understanding pages: readability",
   "guide": null,
   "modified": "2026-10-09"
  },
  {
   "id": "server-side-rendering-and-static-generation",
   "url": "https://ghostagentlab.com/glossary/server-side-rendering-and-static-generation/",
   "name": "Server-side rendering (SSR) and static generation",
   "definition": "Ways of building pages so the content is already in the raw HTML, rather than added afterwards by JavaScript. See why AI agents can't see JavaScript-only content.",
   "why": "How your pages are built decides whether agents see your products or an empty page. Sites that add their content with JavaScript can look perfect to shoppers while appearing blank to many AI crawlers and assistants. Server-side rendering or static generation fixes that, and usually helps search engines and page speed too. It's a decision for your development team or platform.",
   "example": "A store built as a single-page app turns on server rendering for its product pages, so name, price, availability and shipping arrive in the HTML, while personalized recommendations and the chat widget still load with JavaScript.",
   "howToCheck": "A free AgentScore scan compares the text in your HTML with the text after JavaScript runs, and passes when at least 70% is already in the HTML. For a single page, What does ChatGPT see? at /tools/ai-view/ shows it without JavaScript next to your live page.",
   "related": [
    "raw-html",
    "structured-data",
    "ai-search-crawler",
    "agent-readiness"
   ],
   "section": "Understanding pages: readability",
   "guide": "https://ghostagentlab.com/articles/javascript-content-ai-agents/",
   "modified": "2026-10-09"
  },
  {
   "id": "structured-data",
   "url": "https://ghostagentlab.com/glossary/structured-data/",
   "name": "Structured data (schema.org, JSON-LD)",
   "definition": "Machine-readable facts in your page code, such as a product's name, price and stock, written in the shared schema.org vocabulary, usually as JSON-LD. Agents can read it without guessing. See product data AI shopping agents can read.",
   "why": "Structured data is how agents read your prices, stock and product details without guessing. In our study of 255 online stores, 63% gave price, currency and availability this way, and 19% had no usable product data at all. When agents can read these facts reliably, they can compare you fairly and quote shoppers the right price.",
   "example": "A product page shows a sale price, a bundle price and a \"from\" price side by side, but its JSON-LD states \"price\": \"18.00\", \"priceCurrency\": \"USD\" and InStock availability, so an agent quotes the right one.",
   "howToCheck": "Paste a product page address into the Schema Markup Validator to check the markup, and Google's Rich Results Test to see which fields Google can use. A free AgentScore scan checks your structured data, product price and stock, and whether they agree with what the page shows.",
   "related": [
    "raw-html",
    "canonical-url",
    "agentic-commerce",
    "mcp"
   ],
   "section": "Understanding pages: readability",
   "guide": "https://ghostagentlab.com/articles/structured-data-ai-shopping-agents/",
   "modified": "2026-10-09"
  },
  {
   "id": "xml-sitemap",
   "url": "https://ghostagentlab.com/glossary/xml-sitemap/",
   "name": "XML sitemap",
   "definition": "A file listing the pages you want found, so crawlers and agents don't depend on following links. See XML sitemaps for AI agents.",
   "why": "A sitemap makes sure agents can find every page you want found, not just the ones linked from your home page. That matters most for large catalogues, where new or deep product pages are easy to miss. Keep it complete and up to date, and list it in robots.txt so agents know where it is.",
   "example": "A store's robots.txt says Sitemap: https://northwind.example/sitemap.xml, which points to an index listing separate sitemaps for products, collections and pages, so an agent can reach a deep product page without clicking through menus.",
   "howToCheck": "Open your robots.txt, look for a Sitemap line, and confirm the file it points to loads as a list of your pages rather than your home page. AgentScore's \"Valid sitemap\" check looks there and at /sitemap.xml and /sitemap_index.xml, and reports how many entries it found.",
   "related": [
    "robots-txt",
    "canonical-url",
    "llms-txt",
    "check"
   ],
   "section": "Understanding pages: readability",
   "guide": "https://ghostagentlab.com/articles/xml-sitemaps-ai-agents/",
   "modified": "2026-10-09"
  },
  {
   "id": "canonical-url",
   "url": "https://ghostagentlab.com/glossary/canonical-url/",
   "name": "Canonical URL",
   "definition": "The one address you name as the main version of a page that can be reached in several ways. It stops agents treating duplicates as different products. See canonical URLs.",
   "why": "Most online stores can show the same product at several addresses, through filters, tracking tags or sale categories. Without a canonical URL, agents may treat these as different products, split what they know between them or quote an outdated version. Naming one main address keeps agents, and search engines, focused on the right page.",
   "example": "The same rain jacket appears at /products/rain-jacket?utm_source=newsletter, /products/rain-jacket?variant=4417 and /sale/rain-jacket, and each one names /products/rain-jacket as its canonical URL.",
   "howToCheck": "Open a product page with ?utm_source=test on the end, view the source, and check the canonical link points to the clean address. Then spot-check ten addresses from your sitemap: each should load normally and name itself as canonical.",
   "related": [
    "xml-sitemap",
    "structured-data",
    "raw-html"
   ],
   "section": "Understanding pages: readability",
   "guide": "https://ghostagentlab.com/articles/canonical-urls-ai-agents/",
   "modified": "2026-10-09"
  },
  {
   "id": "alt-text",
   "url": "https://ghostagentlab.com/glossary/alt-text/",
   "name": "Alt text",
   "definition": "A short text description of an image. Agents can't see pictures, so this is how they learn what a product photo shows. See alt text for AI agents.",
   "why": "Agents can't see your photos, so alt text is how they learn what a product looks like, which colour it comes in or what's in the box. Good alt text helps agents describe and compare your products accurately, and it's already expected for accessibility. Your content or merchandising team can usually add it in the tools they use for product pages.",
   "example": "A product photo saved as IMG_4471_final.jpg tells an agent nothing, while alt=\"Northwind waxed cotton field jacket in navy, front view\" tells it the product and the color.",
   "howToCheck": "A free AgentScore scan runs \"Images have text descriptions\": it passes when 90% or more of the visible images on a page have alt text. It can't judge whether descriptions are good, so ask your content team to read the alt text on your bestsellers.",
   "related": [
    "accessible-name",
    "wcag",
    "structured-data"
   ],
   "section": "Understanding pages: readability",
   "guide": "https://ghostagentlab.com/articles/image-alt-text-ai-agents/",
   "modified": "2026-10-09"
  },
  {
   "id": "accessibility-tree",
   "url": "https://ghostagentlab.com/glossary/accessibility-tree/",
   "name": "Accessibility tree",
   "definition": "The structured view of a page that screen readers and many browser agents use: its headings, links, buttons and form fields, with their names. If something isn't in it, many agents can't use it. See what an AI agent sees.",
   "why": "Many browser agents decide what to click by reading the accessibility tree, the same view screen readers use. A button or field that isn't properly described there is invisible to them, even though shoppers can see it perfectly well. Fixing it helps customers using assistive technology and agents at the same time, so the work pays twice.",
   "example": "In a fictional store's accessibility tree, \"Add to cart\" is a named button, but the cart icon shows only as button with no name and the sizes appear as generic boxes, so an agent can't find the cart or pick a size.",
   "howToCheck": "In Chrome's developer tools, the Accessibility pane shows each element's role and name; look for buttons and links with no name. A quick non-technical test is to tab through your header with the keyboard: if you can't reach the menu or the cart, neither can many agents.",
   "related": [
    "accessible-name",
    "browser-agent",
    "wcag",
    "raw-html"
   ],
   "section": "Finding the way: navigability",
   "guide": "https://ghostagentlab.com/blog/what-an-ai-agent-sees/",
   "modified": "2026-10-09"
  },
  {
   "id": "accessible-name",
   "url": "https://ghostagentlab.com/glossary/accessible-name/",
   "name": "Accessible name",
   "definition": "The name a button, link or field has in the accessibility tree, from its text, label or aria-label. An icon-only button with no name is invisible to agents.",
   "why": "Unnamed buttons are the most common problem we find: in our study of 255 online stores, 58% had buttons or links with no name an agent can read, most often icon-only cart, search and menu buttons. An agent can't press what it can't identify. Naming them is usually quick work for a web developer, and it helps screen reader users too.",
   "example": "A header cart icon given aria-label=\"Cart, 2 items\" reads to an agent as a cart button with two items in it. The same icon with no label reads only as \"button\".",
   "howToCheck": "A free AgentScore scan runs \"Buttons and links have names\" and says how many have none. To look at a single button, open Chrome's developer tools: the Accessibility pane shows its name, and an empty name is the problem.",
   "related": [
    "accessibility-tree",
    "overlay",
    "wcag",
    "browser-agent"
   ],
   "section": "Finding the way: navigability",
   "guide": null,
   "modified": "2026-10-09"
  },
  {
   "id": "landmarks",
   "url": "https://ghostagentlab.com/glossary/landmarks/",
   "name": "Landmarks",
   "definition": "Labels for the main regions of a page, such as header, navigation, main content and footer, that let agents skip straight to what they need.",
   "why": "Landmarks let an agent jump straight to your main content instead of reading every menu and banner first, which makes it faster and less likely to give up. They also help people using screen readers. Adding them is usually a small template change for your developers, done once for the whole site.",
   "example": "A store puts its breadcrumb trail in a labeled navigation landmark and its product details in the main landmark, so an agent landing on a product page skips the header menu and promo banner and reads the product first.",
   "howToCheck": "A free AgentScore scan runs \"Page has main and navigation landmarks\" on each page it renders: both present passes, one missing warns and both missing fails. If it flags a gap, ask your developer to add them to the page templates.",
   "related": [
    "accessibility-tree",
    "wcag",
    "browser-agent"
   ],
   "section": "Finding the way: navigability",
   "guide": null,
   "modified": "2026-10-09"
  },
  {
   "id": "overlay",
   "url": "https://ghostagentlab.com/glossary/overlay/",
   "name": "Overlay",
   "definition": "Anything that covers the page, such as a cookie banner, newsletter pop-up or chat window. If an agent can't close it, the journey ends there. See cookie banners and pop-ups.",
   "why": "Pop-ups and banners end more agent journeys than almost anything else, because an agent that can't close one can't reach the page underneath. In our study of 255 online stores, 38% showed one that covered much of the page on arrival, and 24% had no clearly labelled button to close it. Give every overlay a clearly named close button, and keep it clear of buy and checkout buttons.",
   "example": "A cookie banner offers \"OK\" and \"Customize\" with a small unlabeled ×, and on a phone-sized window the chat launcher sits on top of \"Add to cart\", so an agent's click lands on the wrong thing.",
   "howToCheck": "Open your site in a private window and try to get past every pop-up with the keyboard alone, then check a product page and the cart at a narrow window size. A free AgentScore scan reports any overlay covering the page, how much of the screen it covers and which buttons it offers.",
   "related": [
    "accessible-name",
    "browser-agent",
    "flaky",
    "captcha-or-challenge"
   ],
   "section": "Finding the way: navigability",
   "guide": "https://ghostagentlab.com/articles/cookie-banners-popups-ai-agents/",
   "modified": "2026-10-09"
  },
  {
   "id": "wcag",
   "url": "https://ghostagentlab.com/glossary/wcag/",
   "name": "WCAG",
   "definition": "The Web Content Accessibility Guidelines, the standard for making sites usable by people with disabilities. Much of it helps agents too. See accessibility work is agent readiness work.",
   "why": "WCAG work and agent readiness overlap heavily: clear labels, real buttons and well-structured pages help people with disabilities and AI agents alike. If you already invest in accessibility, you're partway to being agent-ready, and the business case for both gets stronger together. Many agent fixes can be planned as part of existing accessibility work.",
   "example": "WCAG 2.2 criterion 4.1.2 Name, Role, Value lines up with AgentScore's \"Buttons and links have names\" check, so labeling an icon-only cart button fixes an accessibility issue and an agent issue at once.",
   "howToCheck": "Ask your accessibility lead or agency for open issues under criteria such as 4.1.2, 1.1.1 and 2.4.1, which match AgentScore checks for named controls, alt text and landmarks. A free AgentScore scan shows where those checks stand from an agent's point of view.",
   "related": [
    "accessibility-tree",
    "alt-text",
    "landmarks",
    "agent-readiness"
   ],
   "section": "Finding the way: navigability",
   "guide": "https://ghostagentlab.com/blog/accessibility-is-agent-readiness/",
   "modified": "2026-10-09"
  },
  {
   "id": "journey",
   "url": "https://ghostagentlab.com/glossary/journey/",
   "name": "Journey",
   "definition": "A task a visitor sets out to complete on your site, such as finding a product and reaching checkout, or booking a call. The journeys that make you money are the ones to test first.",
   "why": "Journeys are where agent readiness turns into revenue: an agent that reads your pages but can't finish checkout still loses you the sale. Start with the two or three journeys that make you money, such as product to checkout or booking a call, and test those first. Those are the ones worth watching every day.",
   "example": "A store writes its key journey the way a customer would ask an assistant: \"Find a waterproof jacket in medium under $150 and add it to the cart.\" A specific goal gives a clearer result than \"buy something\".",
   "howToCheck": "List your two or three money journeys, then give each to an AI assistant's agent mode on your site, telling it not to buy anything. Run each at least three times and note every hesitation, wrong click and retry.",
   "related": [
    "task-completion",
    "browser-agent",
    "ghost-agent",
    "guest-checkout"
   ],
   "section": "Getting it done: tasks and commerce",
   "guide": null,
   "modified": "2026-10-09"
  },
  {
   "id": "task-completion",
   "url": "https://ghostagentlab.com/glossary/task-completion/",
   "name": "Task completion",
   "definition": "Whether an agent can actually finish a journey. It's the AgentScore category worth the most points, and the only way to measure it is to try. See how to test your key journeys.",
   "why": "Task completion is the part of agent readiness that maps most directly to revenue: can an agent actually finish what the customer asked it to do? It's worth the most AgentScore points because a site can be readable and still fail at the last step. The only way to know is to send a real agent through your key journeys and watch what happens.",
   "example": "A store passes the checks for labeled fields, named buttons and structured data, yet agents still fail because its custom size picker can't be operated and the cookie banner reappears on the cart page.",
   "howToCheck": "Send an AI assistant through your key journey and ask it what went wrong if it fails: the answer often names the exact control. To keep watching, a Ghost Agent runs the journey on a schedule and shows the step where it got stuck.",
   "related": [
    "journey",
    "ghost-agent",
    "overlay",
    "guest-checkout"
   ],
   "section": "Getting it done: tasks and commerce",
   "guide": "https://ghostagentlab.com/articles/test-journeys-with-ai-agents/",
   "modified": "2026-10-09"
  },
  {
   "id": "mcp",
   "url": "https://ghostagentlab.com/glossary/mcp/",
   "name": "MCP (Model Context Protocol)",
   "definition": "A standard way to give AI agents direct tools for your site, such as search, product details and cart, instead of making them click through pages.",
   "why": "MCP lets an AI agent use your site through tools you choose, such as search, product details and cart, instead of clicking through pages built for people. That can make agents faster and more reliable on your site. It's early: in our study of 255 online stores, 27% offered an MCP server or WebMCP tools. Ask your platform provider whether they support it.",
   "example": "Instead of finding the search box and working out the size picker, an agent calls a store's search_products tool with \"waterproof jacket, medium\" and gets back a clean list of products, prices and stock.",
   "howToCheck": "Ask your commerce platform whether it offers an MCP server for your store. A free AgentScore scan looks for one at /.well-known/mcp.json or /.well-known/mcp, in a link on your home page, or listed in your llms.txt.",
   "related": [
    "webmcp",
    "agentic-commerce",
    "llms-txt",
    "structured-data"
   ],
   "section": "Getting it done: tasks and commerce",
   "guide": null,
   "modified": "2026-10-09"
  },
  {
   "id": "webmcp",
   "url": "https://ghostagentlab.com/glossary/webmcp/",
   "name": "WebMCP",
   "definition": "A proposal for offering those same tools from inside your web pages, so a browser agent can call them while it's on your site. See MCP and WebMCP.",
   "why": "WebMCP would let browser agents use the same reliable tools while they're on your pages, instead of guessing which buttons to press. It's still a proposal, so treat it as something to watch rather than build for today. If your platform or developers already offer MCP tools, keep an eye on how this develops.",
   "example": "A store's product page registers a search_products tool through navigator.modelContext, reusing the code its search box already calls, so a browser agent can search the catalog without typing into the box.",
   "howToCheck": "A free AgentScore scan looks for WebMCP tools registered on your home page and key pages, such as product pages, and for forms marked with a toolname attribute, as part of its MCP check. Since browsers don't ship WebMCP by default yet, ask your developers whether it's on their watch list rather than their build list.",
   "related": [
    "mcp",
    "browser-agent",
    "agentic-commerce"
   ],
   "section": "Getting it done: tasks and commerce",
   "guide": "https://ghostagentlab.com/articles/mcp-webmcp-for-websites/",
   "modified": "2026-10-09"
  },
  {
   "id": "agentic-commerce",
   "url": "https://ghostagentlab.com/glossary/agentic-commerce/",
   "name": "Agentic commerce",
   "definition": "Purchases that an AI agent makes for a shopper. Protocols such as the Agentic Commerce Protocol, the Universal Commerce Protocol and AP2 aim to make that safe for buyer, store and payment provider. See agentic commerce protocols explained.",
   "why": "Agentic commerce is where agents stop recommending and start buying, so it's the part of agent readiness closest to revenue. The protocols are young, but in our study of 255 online stores 27% already published an agentic checkout profile. Ask your commerce platform and payment provider which protocols they support, and what you'd need to switch on.",
   "example": "A store whose platform publishes a Universal Commerce Protocol profile at /.well-known/ucp passes AgentScore's agentic commerce check. A store that supports only ACP, which is set up privately with each AI platform, shows a warning because there's nothing on its site for a scanner to find.",
   "howToCheck": "Ask your commerce platform and payment provider which protocols they support today, whether it's on by default, and how agent orders are marked in your order system. A free AgentScore scan checks whether your store publishes a UCP profile.",
   "related": [
    "mcp",
    "guest-checkout",
    "structured-data",
    "task-completion"
   ],
   "section": "Getting it done: tasks and commerce",
   "guide": "https://ghostagentlab.com/articles/agentic-commerce-protocols/",
   "modified": "2026-10-09"
  },
  {
   "id": "agent-readiness",
   "url": "https://ghostagentlab.com/glossary/agent-readiness/",
   "name": "Agent readiness",
   "definition": "How well AI agents can get into your site, understand it, find their way around it and complete tasks on it. See what is agent readiness?",
   "why": "Agent readiness is whether your site still works when the visitor is an AI agent acting for a customer. As more research and buying happens through assistants, a site that agents can't use loses sales it never sees. It gives marketing, e-commerce and development a shared way to find and fix the problems, and to show progress.",
   "example": "A store ranks well in Google, which runs JavaScript, yet its prices load with JavaScript that many AI assistants never run, so the same product page is complete for Google and nearly blank to the assistants.",
   "howToCheck": "Work through the 30-minute starter checklist in what is agent readiness?: open /robots.txt, ask how your bot protection treats verified AI agents, view the source of a product page to find the price, and tab through your header. Then run a free AgentScore scan at /agentscore/ and fix the top three items it lists.",
   "related": [
    "agentscore",
    "robots-txt",
    "raw-html",
    "task-completion"
   ],
   "section": "Measuring it",
   "guide": "https://ghostagentlab.com/articles/agent-readiness/",
   "modified": "2026-10-09"
  },
  {
   "id": "agentscore",
   "url": "https://ghostagentlab.com/glossary/agentscore/",
   "name": "AgentScore",
   "definition": "Our free score out of 100 for agent readiness, across Access (25 points), Readability (25), Navigability (20) and Task completion (30). The free scan marks Task completion “not tested” and scores the other three. See how AgentScore works.",
   "why": "AgentScore gives you one number to track agent readiness over time and to compare with others: in our study of 255 online stores, the median was 80. Each point lost comes with a specific finding and a fix, ranked so your team knows where to start. Re-scan after you ship changes to see the score move.",
   "example": "A free scan scores a site 72 when Task completion isn't tested, meaning 72% of the points that could be tested. If AI agents are blocked, the score can't go above 59 until that one failure is fixed.",
   "howToCheck": "Enter your website and a work email at that domain at /agentscore/ to get your score and every fix, with the full report by email. Re-scan after you ship changes.",
   "related": [
    "check",
    "agent-readiness",
    "task-completion",
    "bot-protection"
   ],
   "section": "Measuring it",
   "guide": "https://ghostagentlab.com/blog/how-agentscore-works/",
   "modified": "2026-10-09"
  },
  {
   "id": "check",
   "url": "https://ghostagentlab.com/glossary/check/",
   "name": "Check",
   "definition": "One thing AgentScore tests, such as “Buttons and links have names”, reported as a pass, warning or fail, with who usually fixes it.",
   "why": "Checks turn agent readiness into a to-do list. Each one says what passed or failed, why it matters, and who usually fixes it, so you can hand it to the right team without translating technical findings first. Start with the failures that block agents completely, then the warnings.",
   "example": "\"Buttons and links have names\" might fail with a summary of how many buttons have no name, marked Critical because a failed check with weight 4 or more is Critical, and owned by a developer.",
   "howToCheck": "In your AgentScore report, open \"Why it matters and how to fix it\" under each finding to see the fix and who usually owns it. In the app, Export CSV lists every finding, so you can sort by \"Usually fixed by\" and send each owner their own list.",
   "related": [
    "agentscore",
    "accessible-name",
    "robots-txt"
   ],
   "section": "Measuring it",
   "guide": null,
   "modified": "2026-10-09"
  },
  {
   "id": "ghost-agent",
   "url": "https://ghostagentlab.com/glossary/ghost-agent/",
   "name": "Ghost Agent",
   "definition": "A real AI agent we send through a journey on your site, on a schedule, to see whether agents can complete it. You get a step-by-step replay of each run, and an alert when a journey breaks. See Ghost Agent.",
   "why": "A Ghost Agent tells you whether agents can actually finish your key journeys, not just whether your pages look ready. Because it runs on a schedule, it catches the day a release, a new pop-up or a bot rule breaks checkout for agents, before it costs you weeks of sales. The replay shows exactly where it got stuck.",
   "example": "A store sets up a checkout test that starts on a product page and passes when the agent reaches the payment step. When a run fails, the replay opens on the step where it got stuck and the team gets an email or Slack alert.",
   "howToCheck": "Start a 30-day trial, add your site and set up a Ghost Agent test for your most valuable journey, run by hand, daily or hourly. To spot Ghost Agent visits in your logs, look for GhostAgent/1.0 at the end of the user agent.",
   "related": [
    "journey",
    "run-and-pass-rate",
    "web-bot-auth",
    "task-completion"
   ],
   "section": "Measuring it",
   "guide": "https://ghostagentlab.com/ghost-agent/",
   "modified": "2026-10-09"
  },
  {
   "id": "run-and-pass-rate",
   "url": "https://ghostagentlab.com/glossary/run-and-pass-rate/",
   "name": "Run and pass rate",
   "definition": "A run is one attempt by one agent at a journey. AI agents don't behave the same way every time, so journeys are tried several times, and the pass rate is the share of runs that succeed.",
   "why": "Because AI agents don't do things exactly the same way twice, one successful try doesn't prove much. A pass rate across several runs tells you how reliably agents can complete a journey, which is what matters for revenue. Watch it over time: a falling pass rate is an early warning that something on the site has changed.",
   "example": "A checkout journey is run three times and passes twice, a pass rate of two in three, which means about one customer's assistant in three fails.",
   "howToCheck": "When testing by hand, repeat each journey at least three times before drawing conclusions. Ghost Agents run each check several times and report it as passing, flaky or failing, so you can watch the pass rate over time.",
   "related": [
    "flaky",
    "ghost-agent",
    "journey"
   ],
   "section": "Measuring it",
   "guide": null,
   "modified": "2026-10-09"
  },
  {
   "id": "flaky",
   "url": "https://ghostagentlab.com/glossary/flaky/",
   "name": "Flaky",
   "definition": "A journey where some runs pass and some fail. Agents can do it, but not reliably, which usually points to a slow page, a pop-up that appears sometimes, or an unclear control.",
   "why": "A flaky journey means some customers' agents succeed and others quietly fail, which is easy to miss and hard to explain in the numbers. It usually points to something fixable, such as a slow page, a pop-up that appears only sometimes, or a button that's hard to identify. The run replays show which.",
   "example": "For example, a store's add-to-cart test passes on most runs but fails whenever a newsletter pop-up happens to appear first and covers the button.",
   "howToCheck": "Compare a failed run with one that passed and look for what was different, such as a pop-up, a slow page or a control the agent couldn't identify. Ghost Agent replays open on the step where the run got stuck, next to the last run that passed.",
   "related": [
    "run-and-pass-rate",
    "overlay",
    "accessible-name",
    "ghost-agent"
   ],
   "section": "Measuring it",
   "guide": null,
   "modified": "2026-10-09"
  },
  {
   "id": "agent-traffic",
   "url": "https://ghostagentlab.com/glossary/agent-traffic/",
   "name": "Agent traffic",
   "definition": "Visits from AI agents of every kind. Most never run your analytics tag, so it's measured from server or CDN logs. See how to measure agent traffic.",
   "why": "Agent traffic shows how much of your audience is now software acting for people, and most of it never appears in your analytics. Measuring it tells you which agents visit, what they read and whether they get what they came for. It's the baseline for every decision about letting agents in, blocking them or improving what they see.",
   "example": "A CDN log shows ChatGPT-User asking for /products/trail-shoe and getting a 403: a customer's question went unanswered, and nothing about it appeared in the store's analytics.",
   "howToCheck": "Ask whoever runs your hosting for CDN or server logs, such as Cloudflare Logpush or a Vercel log drain, then count requests by kind of agent and check how many got 403, 429 or 404. Hosted platforms like Shopify don't usually give raw logs, so ask yours what bot reporting it offers.",
   "related": [
    "user-agent",
    "verification",
    "ai-referral",
    "cdn"
   ],
   "section": "Measuring it",
   "guide": "https://ghostagentlab.com/articles/measure-ai-agent-traffic/",
   "modified": "2026-10-09"
  },
  {
   "id": "ai-referral",
   "url": "https://ghostagentlab.com/glossary/ai-referral/",
   "name": "AI referral",
   "definition": "A person who arrives at your site by clicking a link in an AI assistant's answer. Unlike agent visits, these do show up in analytics. See tracking visits and sales from AI assistants.",
   "why": "AI referrals are the part of AI that already shows up in your analytics and sales: people who clicked through from an assistant's answer. Tracking them puts a number on what AI is worth to your business today, and shows which assistants send you customers. Pair them with agent traffic to see what agents read for every visitor they send.",
   "example": "A shopper clicks a link in a ChatGPT answer that carries utm_source=chatgpt.com and lands on a product page, so the visit shows in analytics with chatgpt.com as its source. If the shopper had copied the link into a new tab, it would show as direct.",
   "howToCheck": "In GA4's Traffic acquisition report, change the dimension to Session source and filter for chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and claude.ai. For ongoing reporting, create an \"AI assistants\" custom channel group placed above Referral.",
   "related": [
    "agent-traffic",
    "crawls-per-ai-referral",
    "assistant-fetcher",
    "agentscore"
   ],
   "section": "Measuring it",
   "guide": "https://ghostagentlab.com/articles/ai-referral-traffic/",
   "modified": "2026-10-09"
  },
  {
   "id": "crawls-per-ai-referral",
   "url": "https://ghostagentlab.com/glossary/crawls-per-ai-referral/",
   "name": "Crawls per AI referral",
   "definition": "How many pages an AI company's agents fetched for every visitor its assistant sent you. A low number means you get traffic back for what they read.",
   "why": "This number shows whether an AI company's agents are a fair exchange: a low figure means its assistant sends you visitors for what it reads, and a high one means it reads a lot and sends little back. It's a sound basis for deciding which agents to welcome, limit or block, rather than treating every AI company the same.",
   "example": "For example, if one AI company's agents fetched 2,000 of a store's pages in a month and its assistant sent 100 visitors, that's 20 crawls per AI referral.",
   "howToCheck": "Take one AI company's agent requests from your server or CDN logs, and divide by the visitors its assistant sent you in analytics over the same period. Ghost Agent Lab's Overview shows how many pages each AI company reads for every visitor it sends.",
   "related": [
    "ai-referral",
    "agent-traffic",
    "training-crawler",
    "robots-txt"
   ],
   "section": "Measuring it",
   "guide": null,
   "modified": "2026-10-09"
  }
 ]
}
