Free tool
Is your website
ready for AI agents?
ChatGPT, Claude and other AI agents already read websites every day. Enter your address and we will check your site across 6 areas: what is in order and what needs fixing.
- 6 areas, 18 checks
- Real response data for every result
- Fix examples
- PDF report
Discoverability
Content
Bot access
API, Auth, MCP and Skill discovery
Trust
Commerce
What we check
What the check looks at
Discoverability
3 checks
The basic files an agent needs to find the site and see its pages: robots.txt, a sitemap and Link headers that point to key resources.
- robots.txt file
- sitemap.xml
- Link headers
Content
1 check
Whether the site can serve a page as Markdown instead of HTML when an agent asks for it. This makes reading the content more accurate and cheaper.
- Markdown response
Bot access
3 checks
Separate rules for AI bots in robots.txt, Content Signals that state how the content may be used, and bots proving who they are with a signature.
- AI bot rules in robots.txt
- Content Signals
- Web Bot Auth
API, Auth, MCP and Skill discovery
6 checks
Machine-readable descriptions that let an agent act on the site, not just read it: an API catalog, OAuth discovery, an MCP Server Card, Agent Skills and WebMCP.
- API Catalog
- OAuth / OIDC discovery
- MCP Server Card
- Agent Skills index
- WebMCP
- OAuth Protected Resource Metadata
Trust
1 check
Whether the site opens over HTTPS. It is the minimum requirement for browsers and agents alike.
- HTTPS
Commerce
4 checks
New protocols that let agents pay automatically. Shown for information; they do not affect the total score.
- x402
- MPP (Machine Payment Protocol)
- Universal Commerce Protocol
- ACP (Agentic Commerce Protocol)
Starting
Basic
Intermediate
Advanced
Agent-Native
Levels
How the score works
Every check carries the same weight. Points from passed checks are added up and converted to a 100-point scale. 14 checks count towards the score; the commerce area is for information only.
The score falls into one of 5 levels. A level does not say how “good” a site is, only how clear the signals it gives AI agents are.
Starting
0-19 points
Almost none of the basic signals for agents are there.
Basic
20-39 points
Some basic files are in place, but many gaps remain.
Intermediate
40-59 points
The basic standards are in place; agent capabilities are weak.
Advanced
60-79 points
Most standards are supported; a few steps remain.
Agent-Native
80-100 points
The site is fully open to agents.
Why it matters
Agents read a site through files and rules, not like people do
Showing up in answers
ChatGPT, Claude and other AI assistants read websites themselves when they prepare an answer. If a site gives them no clear signals, its content is less likely to end up in those answers.
From reading to acting
Agents increasingly do more than read: they look things up, compare and place orders. For that, a site needs machine-readable descriptions.
Data, not guesses
The check shows every result with real data: which address was requested and what came back. You do not guess what works, you see it.
Questions
Questions about the check
It shows how easily AI assistants such as ChatGPT and Claude, and their agents, can find, read and understand your website. The check looks at open standards for this: robots.txt, the sitemap, rules for AI bots, machine-readable API and MCP descriptions, and HTTPS.
It is free, with no sign-up or email required. You enter the address and see the result right away; you can also download a PDF report.
It changes nothing. Like an ordinary visitor, we send requests to the home page and to standard addresses (for example /robots.txt, /sitemap.xml and files under /.well-known/). We do not fill in forms, do not log in and do not check internal network addresses.
No. This check does not measure classic SEO. Some of the standards here are new (MCP, Agent Skills, WebMCP) and most websites do not have them yet. A low score means the site gives AI agents few extra signals, not that it ranks poorly in search.
Protocols for agent payments (x402, MPP, ACP and others) are still taking shape and only apply to sites that sell online. So these checks are for information: you see the result, but the total score does not change.
Under every check that did not pass you will find what was missing, why it matters and a fix example. Most of it is server configuration or a few static files. You can do it yourself, or send us the result and we will go through it together.
Area names, levels and explanations are in the language you choose. The technical details of each check (the data found, the description and the fix example) are currently in Azerbaijani.
Let us talk about your project!
