Agent readiness checker
See your website the way AI agents see it. Enter a URL: we check whether AI crawlers may read it, whether agents get the same page as people, what they can understand without JavaScript, and whether they can act on it.
What the checker looks at
AI agents reach your site in different ways: crawlers that build AI search indexes, assistants that fetch a page because someone asked a question, and browser agents that click through it for a user. The checker makes the same requests they make and grades what comes back in four groups.
Discovery25% of the score
Can agents find your pages?
Well-behaved crawlers read robots.txt first. We parse your file the way RFC 9309 describes, with named groups before the catch-all, the longest rule winning, and wildcards and end anchors, and give a verdict for 13 AI crawlers and user agents. AI search and assistant agents are graded apart from training crawlers: blocking OAI-SearchBot or Claude-User keeps you out of answers, while blocking GPTBot or CCBot is a decision about training data that many sites make on purpose. Our guide to robots.txt for AI bots explains which is which.
We also look for a sitemap, listed in robots.txt or at /sitemap.xml, and count its pages, and for llms.txt, a proposed Markdown summary of your site for language models. We check its structure against llmstxt.org: a title, a short summary and sections of links. Support for the file still varies between AI tools (see who reads llms.txt), so a missing file costs only a few points.
Access30% of the score
Do agents get the same page as people?
We request your page twice: once as a regular browser and once as an AI agent that identifies itself as ASDK-AgentReadinessChecker. If the browser gets the page and the agent gets a 403, a CAPTCHA or a challenge screen, a firewall or bot-protection service is turning agents away. Firewalls that verify well-known crawlers by IP address may still let the real ones through, so read this as a strong hint rather than proof. See how to tell when AI agents visit your site.
We also check HTTPS, the time to first byte (measured from our server, so distance adds to it) and noindex, nosnippet and similar directives in meta tags and the X-Robots-Tag header. They keep a page out of AI answers just as they keep it out of search results.
Understanding30% of the score
Can agents read what the page says?
Many AI crawlers do not run JavaScript. We count the words in the HTML your server sends, before any script runs: a page that fills in only after JavaScript looks empty to them. Then we read what tells an agent what the page is about: the title, meta description, canonical URL, language, Open Graph tags and the heading outline.
Structured data in JSON-LD states facts in a form machines don’t have to guess: who you are (Organization, LocalBusiness), what you offer (Product, SoftwareApplication, Service) and common questions (FAQPage). We list every schema.org type we find. Last, we check whether the page offers a Markdown version, through an alternate link of type text/markdown or by answering requests that ask for Markdown. It is an emerging convention that spares agents from parsing your layout.
Actions15% of the score
Can agents do something on your site?
Reading is half the job. Agents that work for a buyer also need to contact you, compare plans or book a call. We look for links to contact, pricing and demo or booking pages, and check that form fields have labels: the same labels screen readers rely on, and the ones agents use to understand what to type.
We also look for WebMCP, a draft browser API that lets a page register its actions as tools an agent can call: tool registration in your scripts (document.modelContext.registerTool) and form annotations (the toolname and tooldescription attributes). As of October 20, 2026, WebMCP is a draft of the W3C Web Machine Learning Community Group, with an origin trial in Chrome from version 149, so a missing setup is a warning, not a failure. Our plain-English guide to WebMCP shows how to start with one form.
How the score works
Each check passes, warns or fails. A category score averages its checks by weight, and the overall score weighs Access and Understanding at 30% each, Discovery at 25% and Actions at 15%. The fix list is sorted by how many points each fix would add.
These are heuristics. We check one page and the files at your site’s root, from one location, at one moment; a firewall, a slow region or a page behind a login can give real agents a different result. For a fuller list of what to change and why, read the agent-ready website checklist.
Questions about the checker
Something else? Write to hello@asdk.com
Is the agent readiness checker free?
Yes. It is free, needs no sign-up and works on any public website. We don’t save the results: the report link simply runs the check again.
Which AI crawlers and agents does it check?
GPTBot, OAI-SearchBot and ChatGPT-User (OpenAI); ClaudeBot, Claude-User and Claude-SearchBot (Anthropic); PerplexityBot and Perplexity-User; Google-Extended; Applebot-Extended; CCBot; Meta-ExternalAgent; and Bytespider.
Does the checker respect robots.txt?
Yes. It identifies itself as ASDK-AgentReadinessChecker and follows your robots.txt rules for that name, or for the * group when there is none. It never requests a page or file on your site that the rules disallow: a disallowed page is skipped, and so is a disallowed llms.txt or sitemap. It sends a handful of requests per check and never crawls your site.
Why does it request my page as a browser too?
To spot firewalls that block AI agents. We fetch the page once with a regular browser user agent and once as our agent; if only the browser gets through, agents are probably being turned away.
What is a good score?
85 or more means agents can find, reach and read the page with only small gaps. Below 60, start at the top of the fix list: it is usually blocked access or content that appears only with JavaScript.
Why can my score differ from other checkers?
Each tool chooses its own checks and weights, and results depend on the page you test and on when and where the requests come from. Use the score to track your own progress, and the fix list to decide what to do next.
Keep your site agent-ready
Agent-Ready is the ASDK agent that makes your site readable and usable for ChatGPT, Gemini, Claude and other AI agents. It is in early access.