Agent readiness check
Scores any public site on the signals AI agents and crawlers actually read — llms.txt, robots.txt, sitemaps, AGENTS.md, markdown negotiation, bot-access signals, protocol discovery and up to 25 pages of metadata. About 20–45 seconds. No sign-up.
Recent scans
live| Host | Level | Score | When |
|---|---|---|---|
| jobsaicopilot.com | L0 | 17/100 | 12d |
| johnlacorte.com | L0 | 17/100 | 2026-08-24 |
| www.herringbickers.com | L1 | 56/100 | 2026-08-24 |
| www.browarddefender.org | L1 | 45/100 | 2026-08-23 |
| www.whosbest.org | L4 | 94/100 | 2026-08-23 |
| example.com | L0 | 28/100 | 2026-08-23 |
| www.wte.net | L1 | 21/100 | 2026-08-23 |
Top scores
all time| Host | Level | Score | Rank |
|---|---|---|---|
| www.whosbest.org | L4 | 94/100 | #1 |
| www.herringbickers.com | L1 | 56/100 | #2 |
| www.browarddefender.org | L1 | 45/100 | #3 |
| example.com | L0 | 28/100 | #4 |
| www.wte.net | L1 | 21/100 | #5 |
| jobsaicopilot.com | L0 | 17/100 | #6 |
| johnlacorte.com | L0 | 17/100 | #7 |
What gets checked
- S1–S15Site-wide — llms.txt exists and is served as text/plain, robots.txt allows AI crawlers, sitemap.xml validity and lastmod, sitemap.md, AGENTS.md sections, HTTPS, root OpenAPI.
- L1–L10llms.txt lint — H1 title, blockquote summary, H2 link sections, link format, sampled links resolve, Optional section, llms-full.txt.
- P1–P24Per page — HTTP 200, redirects, canonical, description, Open Graph, lang, JSON-LD and required fields, headings, text-to-HTML ratio, markdown mirror + frontmatter, alternate/Link headers, content negotiation, JS-render dependency.
- M1–M4Markdown negotiation — Serves text/markdown on request, Vary: Accept, 406 on unsatisfiable Accept, q-values.
- B1–B5Bot access — AI-bot rules, Content-Signal, Link headers on the homepage, DNS-AID records, Web Bot Auth directory.
- C1–C16Protocol discovery — MCP server card, A2A agent card, Agent Skills, API catalog, OAuth metadata, security.txt, OpenAPI, auth.md, UCP/ACP/x402, content parity for GPTBot — informational unless present.
- A1–A23Accessibility — Static-DOM rules on the homepage: alt text, labels, accessible names, heading order, landmarks, lang, viewport zoom, ARIA validity. Separate score.
Agent Readability score = passed ÷ graded over the site-wide and per-page checks (warn counts as a miss; n/a and unreliable are excluded). Bands: Excellent ≥ 90, Good ≥ 70, Fair ≥ 50. Accessibility never affects the main score. The 0–5 level ladder follows the isitagentready model.
How it works
- Enter a URL. Any public site. Pick 5, 10 or 25 pages to sample from its sitemap (the homepage is always included).
- Watch the scan. Progress streams live: page discovery, site checks, llms.txt lint, per-page checks, markdown negotiation, bot access, protocols, accessibility, scoring.
- Fix what matters. The report ranks the next-level requirements, gives a copyable fix and an AI prompt per failed check, and has a permalink plus a Markdown twin you can hand to an agent.
Frequently asked questions
What is the Agent Readability score?
The share of graded site-wide and per-page checks that pass, 0–100, in the style of Vercel's agent-readability audit. Excellent is 90+, Good 70+, Fair 50+. Warnings count as misses; not-applicable and unreliable checks are excluded.
What are the levels?
A 0–5 ladder modelled on isitagentready.com: Level 1 basic crawlability (robots.txt, sitemap.xml or Link headers), Level 2 explicit AI-bot policy and Content-Signal, Level 3 markdown negotiation, Level 4 at least one agent integration (MCP, A2A, Agent Skills or an API catalog), Level 5 auth and trust signals.
Does the scan change or store anything on my site?
No. It only fetches public URLs with the user agent WhosBest-Validator/2.0, respects robots.txt for page sampling, and caps each request at 10 seconds and 2 MB. Results are cached per host for about 10 minutes.
Is it free, and is my result public?
Free, no account. Public scans appear on the leaderboards; tick "Hide from public leaderboards" before scanning, or use "Hide my domain" on the report, to keep a domain off them.
Can I get the result as data?
Every report has a permalink, a Markdown twin (.md) with a what-to-fix-first list, a JSON endpoint and an SVG badge you can embed.