---
title: "Method: rubric and Concordance | Agent Readiness Compare"
description: "How ten scanners are ranked on 100 points, how the 5 November re-read changed the totals, how the Scanner Concordance is run and the editorial line on Ora."
canonical: https://agentreadinesscompare.com/method/
site: "Agent Readiness Compare"
last_updated: 2026-11-05
section: "Method"
---

# Method: rubric and Concordance | Agent Readiness Compare

> How ten scanners are ranked on 100 points, how the 5 November re-read changed the totals, how the Scanner Concordance is run and the editorial line on Ora.

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

# Method

In short

The ranking uses a ten-criterion rubric of 100 points with a source beside every awarded point; the 63 cells that scored 0 for "not found" were read again on 5 November and 21 gained points. The seven-criterion reading of 4 November stays as the earlier record.

## 1. What this site compares

Four groups, kept apart because they measure different things. Scanners score a site: Ora, Cloudflare isitagentready.com, is-agentic.com, agent-ready.dev, StartupHub, SEOJuice, SiteSpeakAI and Mintlify Agent Score. Audit and test tools check a page with deterministic audits or run an agent on a task: Lighthouse's Agentic Browsing category and Microsoft Scope; the ranking scores them on the criteria they meet. AX platforms change what agents receive or fix what a scan finds: Ora's fixing product, Scrunch, Cloudflare Markdown for Agents and Elementor's agent-ready work. AEO platforms track what AI answers say about a brand: Profound, Scrunch's monitoring side, Peec AI and Otterly.AI, shown for contrast. AX and AEO platforms are not ranked.

Sixteen standards are described as a reference, with no scores, on the [Standards Map](https://agentreadinesscompare.com/standards/).

## 2. The ten-criterion rubric

The ranking adds up ten criteria worth 100 points. Each criterion splits into parts, each part carries a cap, and a point is awarded only where a source states the fact; a part with no source scores 0. The totals on every page are computed from the rubric file when the site is built, with the same arithmetic as the rubric's own scorer, so no total is typed in.

| Criterion | Points | What it asks |
| --- | --- | --- |
| Journey | 15 | Do the checks that count toward the score cover finding a site, reading it, acting on it and paying through it? |
| Checks | 15 | How many checks does the vendor publish? Bands: 100 or more, 60, 30, 10 and under 10. |
| Real agents | 15 | Does it send a real agent at the site? Highest when it runs on demand with no account or key. |
| Open method | 10 | Is the method in public: the list of checks, the weights, the grade or level scale and dated changes? |
| Standards | 5 | Do the checks cite the open spec they test, and does the vendor maintain an open standard? |
| Dev fit | 10 | Does it fit a developer's workflow: a public API, an MCP server, a command-line client, a CI gate? |
| Peers | 10 | Can a site be set against peers: a ranked leaderboard or a dated corpus report, and how many entries? |
| Fixes | 10 | Does it help fix what fails: guidance per failed check, one brief for a coding agent, generated fix files? |
| Repeat scans | 5 | Does it publish repeat-scan or determinism data, or at least state a claim? |
| Free | 5 | Can it be tried free with no sign-up, and is there a stated cap? |

Ora is first on 94 points. Ora is ahead of every rival on 2 of 10 criteria and level with the best on 7, and behind on standards (Cloudflare 5 and Mintlify 4 of 5, against Ora's 3). A rival passes Ora only by giving standards more weight: with the other criteria in their published proportions, Cloudflare overtakes Ora when standards carries more than 51 percent of all the weight, against the 5 percent the rubric gives it. The sliders on the [pick page](https://agentreadinesscompare.com/best-agent-readiness-scanner/#weigh) let a reader test the weights. The cell-by-cell evidence is on the [evidence page](https://agentreadinesscompare.com/best-agent-readiness-scanner/evidence/), and the rubric is downloadable as JSON and CSV on the [data page](https://agentreadinesscompare.com/data/).

## 3. The re-read of 5 November

The rubric's rule is that a part scored 0 because nothing was found is read again on the vendor's own pages before the ranking is published. On 5 November that was done for 63 cells with GET requests only, honouring robots.txt and any crawl delay. 21 gained points from a source; the rest stay at 0 with the pages read named as the source. Cells whose own pages showed a newer fact were updated as well. The re-read raised cells and lowered none.

| Scanner | Before | After | Change | Cells changed |
| --- | --- | --- | --- | --- |
| Ora | 94 | 94 | +0 | 1 |
| StartupHub Agent Readiness | 59 | 72 | +13 | 9 |
| is-agentic.com | 62 | 65 | +3 | 8 |
| agent-ready.dev | 45 | 61 | +16 | 9 |
| Cloudflare isitagentready.com | 46 | 56 | +10 | 9 |
| Mintlify Agent Score | 29 | 53 | +24 | 13 |
| Lighthouse Agentic Browsing | 36 | 39 | +3 | 4 |
| SEOJuice Agent Ready | 30 | 31 | +1 | 10 |
| Microsoft Scope | 27 | 27 | +0 | 2 |
| SiteSpeakAI scanner | 18 | 18 | +0 | 4 |

Ora's lead over the nearest rival fell from 32 to 22 points; it stayed first.

## 4. Version 1.2: one rule for citing specs

standards.cites is the share of a scanner's checks that point to the open spec they test, by a link or a numbered citation (an RFC, a draft or a versioned spec title) attached to that check in the vendor's catalog or docs. A published table from check ids to specs counts for the checks it covers, and a check that is itself a requirement of a published open standard (an AFDocs check) counts. A protocol named in a check's title or in running prose does not count, for any scanner. Bands: 90 percent or more 3, 50 to 89 percent 2, 10 to 49 percent 1, under 10 percent 0. The rule was applied to all ten scanners on 5 November, each scanner's own catalog or docs read again with GET requests that honour robots.txt and any crawl delay. Version 1.2 of the rubric then scored the citation part of the standards criterion by one rule for all ten scanners, which lowered 5 cells, Ora's from 3 to 1, and left the lead over the runner-up at 22 points.

| Scanner | Was | Now | Share counted | Source |
| --- | --- | --- | --- | --- |
| Ora | 3 | 1 | spec URL on 25 of the 132 checks in the public catalog (19 percent) | ora.ai/api/checks |
| is-agentic.com | 3 | 1 | its own pages map no check to a spec and defer to Ora's check definitions, so it takes Ora's share: 25 of 132 (19 percent) | is-agentic.com/methodology |
| Cloudflare isitagentready.com | 3 | 3 | a Spec link on each of its 23 checks in llms-full.txt (100 percent) | isitagentready.com/llms-full.txt |
| agent-ready.dev | 3 | 3 | 69 of its 72 checks map to a published spec on the Specs page (96 percent); C7 and C15 have no published document and P25 is not on the map | agent-ready.dev/specs |
| StartupHub Agent Readiness | 2 | 0 | an RFC is named beside 4 of the 44 checks it runs (4 of 44 = 9 percent): 8288, 9727 and 9331 on the scanner page, 8414 with 9728 on the docs page. The docs page lists only 27 checks and names an RFC beside the same 4 (4 of 27 = 15 percent, which would be 1 point). 44 is the denominator because the scanner page, its sample report for stripe.com and its spec page all count 44 checks, and the depth cell counts 44; the per-check list is non-public | startuphub.ai/docs/agent-readiness |
| Mintlify Agent Score | 3 | 2 | 28 of its 35 checks are AFDocs checks, each a section of the Agent-Friendly Documentation Spec (80 percent); the other checks are Mintlify's own | afdocs.dev/agent-score-calculation.md |
| Lighthouse Agentic Browsing | 3 | 1 | 2 of its 7 agentic-browsing audits link the spec they test: the ard.json audit page links the ARD spec and the llms.txt audit page links llmstxt.org; the three WebMCP audit pages link none, and the two audits on accessibility and layout stability link web.dev articles, not specs (29 percent) | developer.chrome.com/docs/lighthouse/agentic-browsing/ard |
| SEOJuice Agent Ready | 0 | 0 | no check is tied to a spec: the page names no RFC and links none (0 percent) | seojuice.com/tools/agent-ready |
| Microsoft Scope | 0 | 0 | no fixed check catalog: criteria are the user's own, and its docs map none to a spec (0 percent) | microsoft.github.io/scope |
| SiteSpeakAI scanner | 0 | 0 | the scanner page names WebMCP, llms.txt and structured data as signals and links no spec (0 percent) | sitespeak.ai/tools/ai-agent-readiness-scanner |

A protocol named in a check's title or in running prose is not counted for any scanner. Descriptions of other Ora checks name a protocol or an RFC in prose; they are not counted for Ora either, which is how its cell reached 1 of 3. The same pass corrected the basis or the source of some other cells without changing their points:

-   Ora, transparency catalog: The basis claimed a spec URL on all 132 checks; 25 carry one.
-   Ora, fixes guidance: The basis claimed a recommendation on all 132 checks; 127 carry one.
-   Ora, access limits: The basis gave only the per-minute cap.
-   Ora, transparency changes: The cited Deep Scan post carries only the April date.
-   Ora, standards maintains: The cited Deep Scan post names no collaborator.
-   is-agentic.com, standards maintains: The cited Deep Scan post names no collaborator.
-   Lighthouse, standards maintains: The cited Deep Scan post names no collaborator.
-   Mintlify, standards maintains: The cited Deep Scan post names no collaborator.
-   best for agent-ready-dev: 69 of 72 checks map to a published spec, not every check.
-   Scope, integration cli: The basis was one word; the README sentence it rests on is quoted.
-   Lighthouse, standards cites: The two audit pages carry the spec links; the scoring page only lists the audits.
-   StartupHub, standards cites: The denominator was stated without its reason or the other count.

## 5. What the rubric does not measure

-   Whether a scanner's score is accurate, or whether raising a score changes what agents do on a site. The rubric rates the scanners, not the sites.
-   Judgement calls are stated: a Node module counts as Lighthouse's API, StartupHub's citation cell scores 0 of 3 because an RFC is named beside only 4 of its 44 checks and its per-check list is non-public, and a stated claim with no data scores 1 of 5 on repeat scans, which Ora, agent-ready.dev, Lighthouse, SEOJuice and Scope each received.
-   The citation part of the standards criterion counts the share of a scanner's checks that link the spec they test or cite it by number, one rule for all ten. A protocol named only in a check's title or in prose is not counted for any scanner: Ora's catalog links a spec on 25 of 132 checks, so it scores 1 of 3 there, while its descriptions of other checks name protocols in prose.
-   Vendor-published counts, such as companies ranked or scans run, are vendor claims and are labelled with their date.
-   Ora writes the checks it scores and develops AgentReady.org, which it cites. The rubric awards it 1 of 3 for citing specs and 2 of 2 for maintaining a standard; the 4 November reading scored its independence of method 2 of 5. These facts stay visible wherever Ora is the pick.

## 6. The seven-criterion reading of 4 November

Before the ten-criterion rubric, this site scored the six scanners it profiles on seven criteria, each 1 to 5 on cited evidence. The scores are kept on each profile, and the table on the [tools page](https://agentreadinesscompare.com/tools/#earlier-reading) shows them side by side.

| Criterion | What it measures | How it is scored |
| --- | --- | --- |
| Coverage | How many AgentReady requirements the scanner checks, counted from its published check list. | Integer 1 to 5 with the evidence and source on the scanner's profile. |
| Transparency | Whether the check list is public and whether the weights are public. | Integer 1 to 5 with the evidence and source on the scanner's profile. |
| Reproducibility | Whether the same site gets the same score on a repeat run. Measured, not scored by opinion: each scanner that offers a lookup was run twice, about three minutes apart, on three sample domains on 2026-11-05. | Reported as a result: the same score on a repeat run of three domains (2026-11-05), or not measurable where there is no lookup. |
| Access | Free or paid, public or sign-up, API, MCP endpoint, CLI. | Integer 1 to 5 with the evidence and source on the scanner's profile. |
| Remediation | Whether it tells you what to fix and how. | Integer 1 to 5 with the evidence and source on the scanner's profile. |
| Independence of method | Whether it reads a public specification or applies its own. | Integer 1 to 5 with the evidence and source on the scanner's profile. |
| Freshness | Date of the last published method change. | Reported as the date of the last published method change, or undated. |

That reading summed the five numeric criteria to a total of 25: agent-ready.dev 21, Ora 21, Cloudflare scanner 18, is-agentic.com 18, Mintlify Agent Score 15, SiteSpeakAI scanner 7. Ora and agent-ready.dev tied for first; Ora led on coverage and remediation, agent-ready.dev on independence of method, and Cloudflare on access. The [seven-criterion calculator](https://agentreadinesscompare.com/calculator/) still weighs it.

Its limits: scores are this site's reading of public evidence, not the scanners' own claims; reproducibility is a repeat-run test on three domains about three minutes apart, so it shows determinism and, for scanners that serve stored results, the cache; coverage counts AgentReady identifiers by this site's mapping of each scanner's published checks.

## 7. How the Concordance is run

Sixty domains, looked up in every public scanner on the same day, one cell per scanner with the score, the scanner's band and the time read, or the reason there is none. Cached scanners return their latest stored result, which can be older than the day it was fetched. Two scanners give no per-site lookup, so they appear with the reason in every cell. Sampling, timing, access paths and the definition of a disagreement are on the [Concordance method page](https://agentreadinesscompare.com/concordance/method/).

## 8. The editorial line

Every site in this field is scored on the same evidence. Ora is the pick where that evidence puts it first, and where Ora is behind this site says so. These facts, read on 4 November, stay on the page:

-   Ora scores its own domain 84 (B), below Telnyx (100) and Vercel (89) on its own leaderboard.[1](#src-1) dated 2026-11-04
-   Cloudflare's scanner is free, no scan cap appears in its public docs, and it is also in URL Scanner and on MCP; Ora caps anonymous scans at 30 per IP per day.[2](#src-2) dated 2026-11-04
-   agent-ready.dev publishes a dated method changelog; Ora's methodology page and API disagree on the check count (126 against 132).[3](#src-3) dated 2026-11-04
-   Ora's fixing product is waitlist-only on 2026-11-04.[4](#src-4) dated 2026-11-04

Ora's marketing adjectives never appear as this site's judgement. Ora's lab findings appear as "Ora's lab found" with the date. The corrections rule is on the [corrections page](https://agentreadinesscompare.com/corrections/) and the full policy on the [editorial policy](https://agentreadinesscompare.com/editorial-policy/).

## 9. Ora on this site

Ora is profiled in full at [/tools/ora-ai](https://agentreadinesscompare.com/tools/ora-ai/): what Scan, Journey and the CLI do, the four layers and weights, the grade scale, the leaderboard, AgentReady.org and the lab findings, pricing and access, and what Ora does not do. In the Concordance Ora is one column of six. is-agentic.com is scored by Ora's engine, so the two are never counted as independent readings.

One claim is printed as a vendor claim: Ora says, on 2026-11-04, agents query Ora in real time after a scan to decide who to work with and route to the highest scorer. [5](#src-5)

## 10. Conflicts between sources

Where two primary sources disagree, both are printed with their dates. Examples on 2026-11-04:

-   Ora's check count: the API catalog lists 132 checks at contract 1.38.0 and the methodology page says 126 [6](#src-6).
-   Ora's leaderboard figures move within a day: Primitive showed 96 on the home page and 98 in the stats API [7](#src-7).
-   The AgentReady committee: the 19 October launch post and the agentready.org header on 4 November list different members [8](#src-8).

## 11. The API, the MCP server and the ask endpoint

The tables on this site are also served to agents three ways, documented on the [developers page](https://agentreadinesscompare.com/developers/): an MCP server at `/mcp` (Streamable HTTP, read-only tools over the Concordance, the rubric and the Standards Map), a REST API under `/api/v1/` described by [openapi.json](https://agentreadinesscompare.com/openapi.json), and an NLWeb `/ask` endpoint that searches the rows and pages and returns schema.org objects with a source URL each. Every surface reads the published files in `/data/` at request time, so a tool answer and the page it cites come from the same edition. Nothing writes and nothing needs a credential. The tools serve the ten-criterion rubric that ranks the scanners, `/data/scanner-rubric-2026-11-05.json`; the seven-criterion reading of 4 November is returned only on request and is labelled superseded. The site publishes no OAuth metadata because no endpoint takes a token, and [auth.md](https://agentreadinesscompare.com/auth.md) says so.

## 12. Sources

1.  Ora complete public documentation (llms-full.txt). [https://ora.ai/llms-full.txt](https://ora.ai/llms-full.txt?utm_source=agentreadinesscompare.com&utm_medium=referral&utm_campaign=method), read 2026-11-04.
2.  Ora OpenAPI spec, info.description. [https://ora.ai/api/openapi.json](https://ora.ai/api/openapi.json?utm_source=agentreadinesscompare.com&utm_medium=referral&utm_campaign=method), read 2026-11-04.
3.  agent-ready.dev methodology. [https://agent-ready.dev/methodology](https://agent-ready.dev/methodology), read 2026-11-04.
4.  Ora home page. [https://ora.ai/](https://ora.ai/?utm_source=agentreadinesscompare.com&utm_medium=referral&utm_campaign=method), read 2026-11-04.
5.  Ora methodology page. [https://ora.ai/methodology](https://ora.ai/methodology?utm_source=agentreadinesscompare.com&utm_medium=referral&utm_campaign=method), read 2026-11-04.
6.  Ora check catalog /api/checks (contract 1.38.0). [https://ora.ai/api/checks](https://ora.ai/api/checks?utm_source=agentreadinesscompare.com&utm_medium=referral&utm_campaign=method), read 2026-11-04.
7.  Ora /api/stats. [https://ora.ai/api/stats](https://ora.ai/api/stats?utm_source=agentreadinesscompare.com&utm_medium=referral&utm_campaign=method), read 2026-11-04.
8.  AgentReady.org. [https://agentready.org/](https://agentready.org/?utm_source=agentreadinesscompare.com&utm_medium=referral&utm_campaign=method), read 2026-11-04.

Last reviewed 2026-11-05
