# How Known-Good grades agent-readiness

> Rubric v1.1 · 14,430 sites listed · placement cannot be bought; tier 3 is a test, not a listing.

Source: https://knowngood.sh/benchmark

## Tiers

- tier 1 — **Declared**: The site says agents are welcome in its robots.txt. A permission, not a capability.
- tier 2 — **Probe-verified**: We fetched the site and tested each signal ourselves. Dated, repeatable, re-checked on a schedule.
- tier 3 — **Behaviourally verified**: An agent completed the action end to end — booked, bought, submitted — and it worked.

## The five probe checks (tier 2)

Each is fetched live; a check passes only on a real response. Since v1.1 markdown negotiation requires a markdown body that differs from the HTML response.

- Markdown negotiation (`Accept: text/markdown`): 10,946 pass
- Markdown twin at a predictable URL (`/page.md`): 5,753 pass
- llms.txt (`/llms.txt`): 11,196 pass
- API catalog (`RFC 9727`): 5,442 pass
- MCP server answers (`/.well-known/mcp/server-card.json → tools/list`): 6,031 pass

## What the numbers do not mean

- 5,089 of the 10,946 negotiating sites are on Mintlify, which does it by default — a platform moving, not that many decisions. Marked "via mintlify" on their reports.
- 769 sites passed on the header but rate-limited the strict re-probe; counted on the header until the next crawl.
- MCP counts only where the server behind the card answered tools/list (or challenged for auth): 6,031 do, 2,830 cards point at nothing. 5,070 are Mintlify docs search, 311 Teachify course catalogues, 642 built their own: https://knowngood.sh/mcp-servers
- No model provider has committed to reading llms.txt in production; we test for it because it is cheap and coherent, not because it is proven to drive citations.

## What is not in the results

Gambling and pharmacy sites are listed on the same terms as everything else when they are real businesses under a real jurisdiction; category finds candidates, hostname and country decide. Flags are reviewed by hand and re-checked every crawl. To dispute one: hello@knowngood.sh with the host.

- `abuse` — excluded (40): pirated streaming, ROM/APK downloads, rogue pharmacies, gambling affiliates and login mirrors, phishing-shaped sites
- `adult` — excluded by default (5): escort and adult directories; the pages still exist and are probed honestly
- `farm / parked` — excluded (25): generated content networks and domainer placeholders that pass every probe
- `platform_subdomain / platform_preview` — listed, ranked last, not indexed (367): trial and preview deployments on a platform's free subdomain (*.mintlify.app etc.); the same product on its own domain ranks normally

## Changelog

- v1.1 · 2026-08-26 — 378 www/apex pairs collapsed to one report each (the host's own redirect or declared canonical decides; 18 no-preference pairs keep both). Alias pages 301; listed count falls accordingly.
- v1.1 · 2026-08-25 — strict markdown negotiation (body tested); 58 header-only passes and ~120 bot-challenge pages removed; platform attribution; parked/farm/abuse/adult exclusion. The September crawl scores v1 and v1.1 side by side.
- v1 · 2026-08 — initial rubric: five probe checks, three tiers.