Get in
Whether your firewall, your rules file and your server let an agent open the site at all, tested under the names ChatGPT, Claude and Perplexity use.
Free tool · No signup
AI Agent Readiness Checker: Is your website ready for AI agents?
AI agents now visit websites for people. They read, compare, fill in forms, book and buy. This free agent readiness checker shows where an agent gets into your website, where it gets stuck and what to fix first.
What it checks
A 0–100 agent readiness score and the step where agents get stuck.
Read
Whether the words are in the page an agent receives, or only appear after code runs in a browser. Page weight, headings and image descriptions.
Understand
Whether an agent can pin down who you are, what you sell, what it costs and how to reach you, and whether a dead link says it is dead.
Act
Whether the next step is a real link, whether form fields and buttons have names, and whether a human check stops the task.
Built for agents
Whether anything on the site is made for agents: a plain summary, a text version on request, a stated choice about AI and a direct way to act.
Who fixes what
Every fix says whether it is one for your developer, one Eli prepares for your approval or one Eli keeps watching.
When to use it
When to test your website for AI agents
- Before a relaunch or a new theme: confirm an agent can still get in, read your pages and reach your main next step.
- After switching on bot protection: check you did not lock out the agents your customers use.
- When AI visitors arrive but do not convert: see whether forms, buttons and next steps work for an agent.
- Before your developer starts: hand over one list of fixes, each with its evidence.
Methodology
How the agent readiness score is calculated
Analysis version 1.0.0 · Scoring methodology 2026-10-04.1 · Updated 2026-10-09
- One bounded visit: your homepage, robots.txt, sitemap and up to seven key pages your own homepage or sitemap links to. Then one request each as ChatGPT-User, Claude-User and Perplexity-User, one request for a text version, one request for a page that does not exist and five standard addresses where sites describe what an agent can do.
- Pages are fetched without running JavaScript, the way agents that fetch pages directly read them. Agents that operate a full browser can run JavaScript and may get further. The homepage is read up to 1.5 MB. If it is cut off there, the checks that depend on the missing part are reported as unable to verify.
- Every check is a fixed rule with a weight. Pass earns full weight, warning half, fail none. Checks that could not be verified are excluded from the score.
- The score is the weighted mean of five steps: Get in 25%, Read 20%, Understand 15%, Act 25%, Built for agents 15%. If every request under an agent’s name is refused, the score cannot be higher than 49, because nothing else on the site can be used. 90 and up reads as agent-ready, 75 to 89 as usable with gaps, 55 to 74 as agents struggle, and below 55 as not ready.
- Built for agents covers conventions that are still emerging, such as a text version on request and a summary file. No AI company has confirmed them as a requirement, which is why they carry the smallest weight and are labelled as emerging.
- We do not send an AI agent to your site, fill in a form or submit anything. “Stuck” means a fixed check failed at that step, not that we watched an agent fail.
If something cannot be tested, the report says “Not tested” or “Unable to verify”. Uncertainty is never turned into a score.
FAQ
AI agent readiness: questions and answers
What is an AI agent readiness checker?
A tool that tests whether AI agents such as ChatGPT, Claude and Perplexity can open your website, read it, understand what you sell and take the next step for a customer. This one scores each of those steps out of 100 and shows the evidence and the fix for every check.
What makes a website agent-ready?
Four things, in this order. Agents are let in instead of blocked. The words are in the page an agent receives, not only visible after code runs in a browser. Your company, your offer, your price and your contact details can be pinned down. And the next step, such as a form, a booking or an order, works without a person guessing. Anything made specially for agents comes after that.
Can AI agents buy from my website?
Only if they can get in, read the page and reach the step where the order or booking starts. This check shows whether those steps are open, for example whether a human check stops the task or whether the button that starts an order is a real link. It does not place an order, so it cannot confirm a completed purchase.
What is an AI agent, and why would one visit my website?
An AI agent is an assistant that does a task for a person instead of only answering a question. ChatGPT, Claude, Perplexity and Gemini can open a website, read it, compare options, fill in a form or start a booking when their user asks. If your site stops them, the task ends there.
Did an AI agent really visit my site?
No. The tool makes plain requests the way an agent does, under each agent’s published name, and runs fixed checks on what comes back. It never fills in or submits a form and never says an agent completed an action.
My site works fine in a browser. Why is the score low?
A browser runs code, loads images and lets a person guess what a button does. An agent often reads the raw page, stops at a size limit and needs every field and button to have a name. A site can look perfect and still be hard for an agent to use.
Do I need everything under Built for agents today?
No. Those are emerging conventions and no AI company has confirmed them as a requirement. Fix what blocks agents first: access, readable pages and a working next step. The agent-specific items are cheap to add and are the smallest part of the score.
Is the AI Agent Readiness Checker free?
Yes. There is no signup, no card and no AI call. You can test any public website, and the result includes the full list of fixes and a brief you can copy for your developer or your coding assistant.
What does Eli do after this check?
This check is one moment in time. Eli connects once, keeps checking who gets in, prepares the fixes it can make for your approval and rechecks after they ship. The fixes only your developer can make stay on one clear list.
The suite
Related free tools
A free check finds the problem. Eli keeps fixing it.
Run the free check first. If the gaps matter, connect Eli and let it keep working on them.
Connect Eli