Agent Readiness Compare

How to brief an agency on agent readiness

Requirements, evidence and acceptance tests a head of web can put in a brief, built on the AgentReady standard.

In short

Put four things in the brief: the AgentReady MUSTs that apply, the pricing requirement, a named scanner with a target, and a re-test date. AgentReady sets no scores, so the target is yours. Use more than one scanner, because Cloudflare does not test JavaScript-only content or status codes.

A brief names the standard, the scanner and the target

AgentReady sets requirements with stable identifiers and sets no scores. A brief therefore needs two more things the standard does not give: a named scanner and a target on it. Without them, "make us agent ready" has no acceptance test.

  1. State which MUSTs apply, and whether AR-ACT-10 applies to your product.
  2. Quote AR-READ-11 if you publish prices.
  3. Name the scanner that judges the work and the target on it.
  4. Require results before and after, each with the scan date.
  5. Name the scanner or check that re-tests each fix.
  6. List what is out of scope, so the agency does not add work the standard does not ask for.

The data supports no particular target. As context, Ora's B band runs from 70 to 85, and on 2026-11-04 6 percent of scanned sites were at B or above, with an average of 43.

The four MUSTs and the pricing requirement belong in scope

Requirements to quote in the brief, from the AgentReady spec
IdentifierRequirementEvidence to ask for
AR-FIND-01Allow agentsA fetch with an agent user agent returns the page and not a 403.
AR-READ-01Answer in initial HTMLThe answer is in the returned HTML without running JavaScript.
AR-READ-02Real status codesA missing page returns an error code, not a soft 404.
AR-ACT-10An MCP server or CLIOnly for products that agents act on. The brief should say whether this applies.
AR-READ-11Pricing as textPricing sits on a first-party page as text, not an image, calculator or sales form.

AR-READ-11 is not a MUST, but it is the item buyers most often leave out. See MUST, SHOULD, MAY for how the levels differ.

Scanners verify some of these and miss others

Cloudflare's scanner does not test JavaScript-only content or status codes, by our mapping of its 23 checks to the AgentReady identifiers. A site can pass it and still fail AR-READ-01 and AR-READ-02. Ora checks whether the answer is in the initial HTML. agent-ready.dev excludes four page checks when a page depends on JavaScript, so a JavaScript-heavy site is not marked down on those four. For AR-READ-01, name a scanner that tests it.

  • Handing work to the agency. Cloudflare provides a prompt per failing check that can go to a coding agent. Vercel's guide walks through each essential with code.
  • Accepting the fix. Ora's POST /api/scan/checks re-runs only the checks you name. Ask the agency to show before and after results with the scan date, since cached scores can move within a day.
  • Reading the result. The Concordance shows scanners disagree by tens of points on the same site, so fix one scanner as the judge in the brief.

Leave out work the evidence does not support

Ora's lab found (19 Oct 2026), over 190 probes on one documentation site, that only JavaScript-only answers and 403 responses stopped retrieval, and that markdown mirrors and negotiation never changed success. These are vendor claims on one site, so treat them as a reason to rank the work, not to drop it. They fit the shape of the MUSTs: the four cover access, initial HTML, status codes and agent actions, and none asks for markdown. A brief that makes markdown mandatory is asking for more than the standard does.

What this guide cannot tell you

  • Whether an agent completes a task on your site. The Vercel guide notes its probes test whether an answer is present in returned content, not whether an agent understands it.
  • Whether the scanner's authors also write the standard it checks against. Ora co-maintains AgentReady and also scores against it; the ten-criterion rubric notes that as a limit, and the 4 November reading scored it 2 of 5 on independence of method. Check the choice against the method page.
  • What the work should cost. The data holds no agency pricing.
  • Whether the standard will change. Version 2.0 added items and retired AR-READ-10 on 2026-10-21.

Sources

  1. AgentReady spec. https://github.com/agentready-org/standard/blob/main/spec/spec.md, read 2026-11-04.
  2. AgentReady changelog. https://github.com/agentready-org/standard/blob/main/CHANGELOG.md, read 2026-11-04.
  3. Cloudflare, Agent Readiness. https://blog.cloudflare.com/agent-readiness/, read 2026-11-04.
  4. Cloudflare, isitagentready llms.txt. https://isitagentready.com/llms.txt, read 2026-11-04.
  5. Ora, llms-full.txt. https://ora.ai/llms-full.txt, read 2026-11-04.
  6. Ora, API stats. https://ora.ai/api/stats, read 2026-11-04.
  7. Ora, methodology. https://ora.ai/methodology, read 2026-11-04.
  8. Ora, AgentReady launch. https://ora.ai/blog/agentready-launch, read 2026-11-04.
  9. agent-ready.dev, methodology. https://agent-ready.dev/methodology, read 2026-11-04.
  10. Vercel, make your site readable by AI agents. https://vercel.com/kb/guide/make-your-site-readable-by-ai-agents, read 2026-11-04.

Last reviewed 2026-11-05