How does AI see your site?

The 25 checks are HTTP-level probes against the open Agent-Adoption Specification — robots.txt, sitemap, llms.txt, markdown content negotiation, OAuth discovery, MCP server cards, and so on. Each returns pass / fail / not-applicable based on what your server actually serves. False positives happen when sites are bot-protected (aggressive WAF rules can block the scan) or behind authentication walls. The signals are real, but their effect on AI visibility is small — agent-readiness is one factor among many.

Yes. No signup, no email, no upsell to see your result. The only limit is 3 scans per minute per IP — same domain twice in a row gets cached so you don't waste quota. The paid CompetLab product covers AI visibility across LLMs, competitors, and traffic — the one-off scan here is free forever.

No. Each scan runs on demand, returns a result, and is gone. The only thing kept in memory is a 1-minute cache of the last few scans (so re-submitting the same domain doesn't re-run the whole probe) — it's per-process, not on disk, and clears within the minute. We don't log domains to a database, don't persist results, and don't share with third parties.

No. In our 908-domain correlation study, only 2 of 50 tested agent-readiness signals reached significance — and none had a 'large' effect. Brand recognition, training data, and content quality matter more in practice. Use the score to spot fixable problems and ship them — don't use it as a ranking lever or a guarantee of AI traffic.

We probe 25 different signals — robots.txt, sitemaps, llms.txt, markdown content negotiation, OAuth endpoints, MCP server cards, and more. Some checks (like rendering strategy) require fetching the page in a real browser engine to see how it actually renders. We trade speed for coverage. Most scans complete in under a minute; complex sites with lots of redirects can take longer. We give the scan up to 5 minutes before timing out.

No. Focus on the scored checks — those are what move your score. Informational checks are useful context (they tell you about emerging signals like AGENTS.md or A2A cards) but don't affect your score. If you want to level up, look at the gate checks for the next level — those are the highest-leverage fixes.

If a check is informational — like AGENTS.md or A2A agent cards — it doesn't move your score, so we don't generate a paste-ready fix-prompt for it. These signals are emerging early-adoption flags, not blockers. Implement them when you're ready to lead on that signal; otherwise, skip them and focus on the scored checks where the fix-prompt appears.

Copy it, paste it into Claude Code, Cursor, or Windsurf. The AI tool reads your codebase and applies the fix. You review the diff, commit. Pure "see → copy → ship" loop — no manual config wrangling, no spec-reading, no deciding what the right header value is. The prompt has all that context built in.

True — most sites don't, today. We track them because adoption is early and the trajectory matters. If you're building a B2B SaaS, having an MCP server card makes your service discoverable as a tool to Claude (and a growing list of other agents). Adding AGENTS.md to a public repo tells AI coding assistants how to work with your codebase. Early movers signal agent-native — and as adoption grows, having these in place pays off.

Yes. The tool works for any public domain. Run three competitors back-to-back to see how their agent-readiness compares to yours. Try the paid CompetLab product to see your full AI visibility picture across competitors and LLMs.

Right after you fix something, to confirm the fix landed. Otherwise, monthly is plenty unless you're shipping infrastructure changes weekly. Don't run it daily to obsess over the score — most of the value is in the failed-check fixes, not the score number itself.