Skip to content
Devtoolry
Sign InJoin for Free

Agent Readiness Checker

Check whether AI agents can use your site: robots.txt rules per AI purpose, bot firewalls, accessible names and labels, llms.txt, Markdown, real 404s and ai-catalog.json.

Something not working, or missing a feature? Send feedback

About the Agent Readiness Checker

The Agent Readiness Checker tests whether AI agents that browse and act for people (ChatGPT agent, Claude, Gemini in Chrome, Perplexity Comet) can reach and use a site. It checks the access policy (robots.txt groups per AI purpose, Content-Signal, and whether a firewall answers bots differently from browsers), whether the content is readable without JavaScript, the accessible names, labels and landmarks agents act on, and the optional discovery files: llms.txt against Lighthouse's rules, Markdown alternates, the sitemap, real 404s and ai-catalog.json.

Our server requests the URLs you enter, like a crawler: it reads the response without running the page's scripts, and the analysis runs in your browser. The URLs and results are not stored.

How to use the Agent Readiness Checker

  1. Enter a page or site address and press Check agent readiness. Pick a page with forms or buttons (search, sign-up, checkout) as well as the home page.
  2. Fix failures under Access policy first: a bot challenge or CAPTCHA on content pages stops agents completely.
  3. Under Accessibility tree, give every icon button an aria-label and every form field a label. Agents find controls by name.
  4. Treat Discovery files and Actions as optional extras, and confirm the result with Lighthouse's Agentic Browsing category in Chrome DevTools.

Common mistakes

  • Content-Signal only in the * group. A crawler obeys only the robots.txt group that names it, so a Content-Signal line under "User-agent: *" does not reach a bot that has its own group. Repeat it in every group.
  • A catch-all 200 page. If every unknown URL returns your home page with 200, agents and Lighthouse read /llms.txt and /.well-known files that do not exist as real. Return 404.

Works well with the Agent Readiness Checker

Frequently asked questions

What does "agent ready" mean?

An AI agent opens pages, reads them and clicks or types for a person. A site is agent ready when agents are allowed in, the content is in the HTML, and every control has a clear name and role in the accessibility tree, which is what agents read. It is mostly good accessibility and a deliberate bot policy.

Is this the same as Lighthouse's Agentic Browsing score?

No. Lighthouse reports a fraction (for example 3/3) from the rendered page, including 33 accessibility rules and layout shift. This checker reads the server HTML and adds checks Lighthouse does not make, such as robots.txt groups and firewall behaviour. Use both.

Do agents follow robots.txt?

Crawlers for training and search do. Fetchers that act for a user (ChatGPT-User, Perplexity-User, Google-Agent) may not, because their operators treat them as the user's own visit. Protect private pages with a login, not robots.txt.

Should I add WebMCP or an ai-catalog.json?

Only when you have something real to offer. WebMCP (tools a page exposes to in-browser agents) is an early proposal in a Chrome origin trial, and ai-catalog.json (Agentic Resource Discovery) is a draft for listing agent resources such as MCP servers. Neither is scored.

Related tools