---
title: "Scanner Concordance method | Agent Readiness Compare"
description: "Sampling, timing, the exact access path per scanner, what each scanner checks and weights, what disagree means and the results of the 2026-11-04 edition."
canonical: https://agentreadinesscompare.com/concordance/method/
site: "Agent Readiness Compare"
last_updated: 2026-11-05
section: "Concordance"
---

# Scanner Concordance method | Agent Readiness Compare

> Sampling, timing, the exact access path per scanner, what each scanner checks and weights, what disagree means and the results of the 2026-11-04 edition.

Source: https://agentreadinesscompare.com/concordance/method/
Site: Agent Readiness Compare
Last reviewed: 2026-11-05

# Concordance method

In short

How the Scanner Concordance is built: sixty domains, 50 from Ora's leaderboard and 10 business sites, each looked up in every public scanner on 4 November 2026. The page gives the exact access path per scanner, what each checks, what disagree means, and the results.

## 1. What this is

Sixty domains, each looked up in every public agent-readiness scanner on the same day (2026-11-04, cells fetched between 19:48 and 19:59 UTC by the machine clock). One row per domain, one cell per scanner: the score, the scanner's own band or grade, and the time the cell was fetched, or the reason the scanner could not give a score. Files are described on the data page.

## 2. Sampling

-   50 domains from Ora's public leaderboard (snapshot the leaderboard snapshot, 3,456 rows, [ora.ai/leaderboard](https://ora.ai/leaderboard?utm_source=agentreadinesscompare.com&utm_medium=referral&utm_campaign=concordance-method)). Six categories: Consumer 8, E-commerce 8, Travel & Hospitality 8, Developer Tools 9, Infrastructure 8, Agent Tools 9. Inside each category the slots are spread round-robin across the grades present (A+, A, B, C, D; the leaderboard holds no F rows on this date, and Consumer and Travel have no A rows), and inside a grade the rows are taken at evenly spaced score positions, preferring rows with a curated brand name. The procedure is deterministic (`build-sample.mjs`).
-   10 mid-sized business sites from the Tranco top 10,000 in the shared crawl universe (the crawl universe, source `tranco`, ranks 3,244 to 8,529), hand-picked across sectors (marketplaces, retail, travel, ticketing, hardware, enterprise software). None is on Ora's leaderboard, so these test scanners on sites nobody tuned for them. Excluded on purpose: CDNs, ad tech, DNS and hosting infrastructure, government, universities, adult and piracy sites.
-   Bias to keep in mind: 50 of 60 rows come from Ora's leaderboard, which lists sites that someone scanned with Ora. That favours Ora and is-agentic.com coverage. The ten Tranco rows are the control.

## 3. Timing

All cells were fetched on 2026-11-04. Each scanner ran on its own lane at one request every 2 to 3 seconds, lanes in parallel. Cached scanners (Ora, is-agentic.com, agent-ready.dev badge, Mintlify) return the latest stored result, which can be older than the fetch: the scan date is kept in `raw.scanned_at` (Ora, is-agentic.com) or `raw.last_computed` (Mintlify) where the scanner exposes it; agent-ready.dev's badge does not expose a scan date. isitagentready.com scanned each site live at fetch time.

## 4. Access method per scanner (exact)

| Scanner | Access we used | Key or sign-up | Notes |
| --- | --- | --- | --- |
| Ora (ora.ai) | GET https://ora.ai/api/score/{domain} (cached score). On a cache miss, POST https://ora.ai/api/scan with {"url":"https://{domain}"}, then poll the GET, capped at 25 fresh scans for the day | none | Documented in ora.ai/llms.txt. Public limit 30 scans a day and 10 a minute per IP. The GET answered HTTP 429 often while the shared crawl was also reading it; after retries, the module falls back to the same-day shared crawl (ora-scores.jsonl, same endpoint) and then to Ora's leaderboard snapshot; raw.via records which. |
| Cloudflare isitagentready.com | POST https://isitagentready.com/api/scan with {"url":"https://{domain}","enabledChecks":[default profile]} | none | Documented for agents in isitagentready.com/llms.txt and /llms-full.txt and listed in its RFC 9727 catalog at /.well-known/api-catalog. No GET-by-domain results endpoint exists. The URL Scanner route (POST /client/v4/accounts/{id}/urlscanner/v2/scan with options.agentReadiness: true, blog.cloudflare.com/agent-readiness) needs a Cloudflare API token, which we did not use. |
| is-agentic.com | GET https://is-agentic.com/api/v1/report?url=https://{domain} (then the www. host) | none | Public, read-only, never starts a scan (https://is-agentic.com/openapi.json). Quota header RateLimit-Policy: "public-report";q=120;w=60. Its llms.txt says automated bulk scanning is not permitted, so domains without a stored report stay unscored; we did not open /scan/{host} to start scans. |
| agent-ready.dev | GET https://agent-ready.dev/api/badge/{domain} and read the SVG aria-label ("agent ready: 71 good" or "unscored") | none | Documented in agent-ready.dev/llms-full.txt; badge cached 24 h. Anonymous POST /api/scan allows 3 scans per 30 days per IP, then asks for an x402 payment; the REST API and MCP need a paid Pro key. We used neither. |
| SiteSpeakAI AI Agent Readiness Scanner | none usable | none for humans | sitespeak.ai/tools/ai-agent-readiness-scanner posts a web form with a CSRF token and session cookie, polls /tools/ai-agent-readiness-scanner/{id}/status and shows /{id}/share. It has no API, no lookup by domain and no public result list. Our rules forbid automating a web form, so every cell is unscored. |
| Mintlify Agent Score | GET https://www.mintlify.com/score/{slug}.md (public markdown; slug = the domain's first label) | none | Scores a documentation site, not a domain. A cell counts only when the scored host is the sample domain or one of its subdomains or paths (for example docs.stripe.com for stripe.com). Leaderboard markdown at mintlify.com/score/index.md; tracker summary at /score/api/agent-rank/summary. |

## 5. What each scanner checks and how it weights (from its own pages)

-   **Ora** (https://ora.ai/methodology): four layers a real agent moves through, Discovery 20 pts (16 checks), Access 30 pts (42 checks), Usability 40 pts (62 checks), Payments 10 pts (6 checks), 126 checks in all. Each layer is scored on checks passed and the layers combine to 0 to 100. Grades: A+ 95 to 100, A 86 to 94, B 70 to 85, C 48 to 69, D 28 to 47, F 0 to 27. Per-domain maxima differ from the page totals (the API returned discovery max 17, accessibility 53, usability 49 for stripe.com), so not-applicable checks appear to be dropped. Check weights change: the ARD check reads "Weighted at 1 point while ecosystem adoption is early (2026-08 audit; was 3)" in the API response.
-   **is-agentic.com** (https://is-agentic.com/methodology): "Every check is run and scored by Ora." It regroups Ora's checks: Essential checks share 80 points, Recommended 20, Emerging signals add a bonus capped at 5; not-applicable checks are excluded and partial results get proportional credit. So is-agentic.com and Ora share a method by design and usually read the same underlying scan (identical `scanned_at`); their numbers still differ because the weighting differs.
-   **Cloudflare isitagentready.com** (https://isitagentready.com/llms-full.txt): 23 checks in 5 categories (Discoverability, Content Accessibility, Bot Access Control, API/Auth/MCP Discovery, Commerce). Two outputs: a readiness level 0 to 5 (gated: Level 1 basic web presence, Level 2 adds AI bot rules and Content Signals, Level 3 adds markdown negotiation, Level 4 adds one of MCP Server Card, A2A Agent Card, Agent Skills, API Catalog, Level 5 adds two of Web Bot Auth, all four integration checks, auth metadata) and a percentage the page computes as passed over counted checks, all weighted equally. Commerce checks are shown but not counted; neutral results are excluded; the default "All Checks" profile leaves out the A2A Agent Card and AP2 checks. Our score column is that page percentage, recomputed from the API response the same way the page script does (read in the page's JavaScript on 2026-11-04); the band column is the level. Colour thresholds on the page: 90 and up good, 50 and up average.
-   **agent-ready.dev** (https://agent-ready.dev/methodology, llms.txt "Last updated: 2026-05-31"; latest logged method change 2026-07-31, accessibility sub-score only): 72 checks against the Vercel Agent Readability Spec, llmstxt.org and agent-protocol specs (MCP, A2A, agents.json and others). Overall score is unweighted: passing checks over total, warnings count as fails, checks marked unreliable by the JS-rendering check are excluded. Bands: 90 to 100 excellent, 70 to 89 good, 50 to 69 fair, 0 to 49 needs improvement. A separate llms.txt sub-score is weighted (structural 3x, content 1x, llms-full.txt 0.5x).
-   **SiteSpeakAI**: the page names WebMCP, llms.txt, structured data, content quality and accessibility, and meta information. It publishes no check count, no weights and no band thresholds.
-   **Mintlify Agent Score** (https://www.mintlify.com/score, built on AFDocs, [afdocs.dev/agent-score-calculation.md](https://afdocs.dev/agent-score-calculation.md)): AFDocs runs 28 checks in 7 categories, weighted Critical 10, High 7, Medium 4, Low 2; warn earns partial credit, skip and not-applicable are excluded; grades A+ 100, A 90 to 99, B 80 to 89, C 70 to 79, D 60 to 69, F 0 to 59. Mintlify's score pages report "Checks passed: x / 35", so Mintlify runs more checks than the 28 AFDocs lists. It scores docs only.

## 6. What disagree means

Scales are all 0 to 100 but they measure different things (Ora: an agent's whole path including use and payment; Cloudflare: presence of 23 standards; agent-ready.dev: 72 spec checks on readability; Mintlify: docs only). Two measures are published:

1.  **Spread** for one domain: the highest minus the lowest score among the independent methods that scored it (Ora, isitagentready.com, agent-ready.dev, Mintlify), counted only when at least three are present. is-agentic.com is left out of the spread because Ora runs and scores its checks; counting it would double-count one method. Its gap to Ora is reported separately in the summary (`ora_vs_is_agentic`).
2.  **Pairwise Spearman rank correlation** for each scanner pair with 15 or more domains scored by both. Spearman compares order, not level, so a scanner that runs 20 points lower than another but ranks sites the same way still correlates near 1.

A large spread does not say which scanner is right. It says the methods look at different things, and the reading for each top disagreement names which checks drive the gap.

## 7. Re-running

The Concordance is re-run monthly with the same sixty domains, replacing a domain only if it disappears, so each edition can show movement. Cells that failed with HTTP 0, 429 or a 5xx are fetched again on the next run. Cached scanners return the latest stored result, which can be older than the fetch.

## 8. Results of this edition (2026-11-04)

-   Scored out of 60: Ora 60 (58 from the cached GET, 2 from fresh POST scans; none needed the crawl or leaderboard fallback in the final pass), isitagentready.com 52 (8 sites refused its fetcher: 7 HTTP 403, 1 HTTP 404), is-agentic.com 47 (13 with no stored report), Mintlify 21 (37 not tracked as docs sites; railway.app and bunnings.com.au rejected because the slug scores docs.railway.com and bunnings.com.ai), agent-ready.dev 0 (every badge read "unscored"; its public corpus page reports 2,570 scanned sites but exposes no per-site lookup), SiteSpeakAI 0 (form only).
-   Medians: Ora 64.5, isitagentready.com 20, is-agentic.com 61, Mintlify 93. The levels differ because the scanners measure different things; read correlations, not levels.
-   Spearman: Ora and isitagentready.com 0.793 (n=52), Ora and is-agentic.com 0.799 (n=47), isitagentready.com and is-agentic.com 0.718 (n=40), Ora and Mintlify 0.466 (n=21), is-agentic.com and Mintlify 0.424 (n=21), isitagentready.com and Mintlify 0.387 (n=18).
-   Ora and is-agentic.com: mean gap is-agentic minus Ora of -2.4 points on 47 rows; only 12 rows read the identical underlying scan (`scanned_at`), because is-agentic.com caches its own copy of an earlier Ora scan.
-   Only 18 rows had three independent scores, so the top-ten disagreement list is mostly Mintlify (docs host) against Cloudflare (main domain, 15 counted standards). With agent-ready.dev and SiteSpeakAI unreadable, the disagreement table is thin; next edition should add a keyed source if one is approved.
-   Cached-score age: several Ora rows on Tranco sites carry scans from September and October 2026 (for example bunnings.com.au scanned 2026-09-03); the cell keeps that date in `raw.scanned_at`.

## 9. Related

-   [The Concordance](https://agentreadinesscompare.com/concordance/)
-   [Edition 2026-11-04](https://agentreadinesscompare.com/concordance/2026-11-04/)

## 10. Sources

1.  Ora methodology page. [https://ora.ai/methodology](https://ora.ai/methodology?utm_source=agentreadinesscompare.com&utm_medium=referral&utm_campaign=concordance-method), read 2026-11-04.
2.  isitagentready.com. [https://isitagentready.com/llms-full.txt](https://isitagentready.com/llms-full.txt), read 2026-11-04.
3.  agent-ready.dev methodology. [https://agent-ready.dev/methodology](https://agent-ready.dev/methodology), read 2026-11-04.
4.  Mintlify Agent Score. [https://www.mintlify.com/score](https://www.mintlify.com/score), read 2026-11-04.
5.  is-agentic methodology. [https://is-agentic.com/methodology](https://is-agentic.com/methodology), read 2026-11-04.

Last reviewed 2026-11-05
