// free tool
Can AI agents actually read your site?
Paste a URL. Get the specific things blocking agents from finding, parsing and using your site, plus the generated files to fix them. No score, no signup, no email.
// glossary
SEO, AEO and GEO: what this tool actually checks
The industry uses these terms loosely, and AEO and GEO in particular get used interchangeably by different writers. Here's the breakdown we build against, and which of this tool's checks belong to each.
Ranking in classic search results (Google, Bing). The oldest discipline here, and still the foundation: titles, meta descriptions, canonical URLs, sitemaps. Everything downstream depends on these being right first.
Getting quoted directly inside an AI-generated answer (ChatGPT, Perplexity, Google AI Overviews) rather than merely ranking a link nearby. This lives or dies on structured data and readable content: an engine can only quote a fact it can parse with confidence.
The broader discipline: making your entire site legible and operable to generative AI systems, not just quotable in one answer. Includes AEO, plus the infrastructure layer: can an agent discover you, fetch you, and (increasingly) transact with you at all.
Questions
What does this actually check?
Eleven core checks: llms.txt, robots.txt access for the major AI crawlers, sitemap, JSON-LD structured data, canonical URL, meta description, OpenGraph tags, title, whether your content survives without JavaScript, OpenAPI discovery and MCP manifest. Turn on commerce mode and it adds eight more for UCP, ACP and AP2.
How is this different from the other agent-readiness scanners?
Most hand you a score. This hands you the files: a generated llms.txt, a JSON-LD block for the page you scanned, a starter UCP manifest, and (if we find an OpenAPI spec) a working MCP server scaffold with typed tools for each operation. The grade is the easy part; the fix is the useful part.
Is the MCP scaffold production-ready?
No, and we won't pretend otherwise. It compiles, runs, and exposes a typed tool per API operation with a documented auth env var. It has no credentials, retry policy, pagination handling or rate limiting. It's a real starting point, not a finished server.
Do you store my scan or my URL?
No. Scans run in memory and nothing is written to a database. There's no signup and no email gate.
Why did some checks come back as 'unknown'?
Unknown means we couldn't fetch something: a timeout, an error, or bot protection blocking us. It never means 'failed'. If a site blocks automated requests, that's worth knowing on its own: AI crawlers often hit the same wall.
// ready_when_you_are
Let's build your agentic advantage
Tell us what should be automated, connected or sold. We'll map the fastest path from idea to a production system.