---
type: Agentic Web Standard
title: "Agent Readability Spec"
description: "How readable and actionable is this site for AI agents from discovery to completion?"
resource: "https://vercel.com/kb/guide/make-your-site-readable-by-ai-agents"
tags: ["read", "document-format", "single-vendor", "serve-content-to-agents", "discover-my-api", "control-ai-crawling"]
status: draft
generated: { by: agenticweb-builder/1.0, at: 2026-10-30T00:00:00Z }
verified: { by: process:agenticweb-engine, at: 2026-10-29T00:00:00Z }
stale_after: 2026-11-28T00:00:00Z
sources:
  - id: ref-1
    resource: "https://vercel.com/kb/guide/make-your-site-readable-by-ai-agents"
    title: "Make your site readable by AI agents — Vercel and Ora research lab — Make your site readable by AI agents (HTML guide, published 28 August 2026, updated 3 September 2026): \"Make your site readable by AI agents\" \"AI agents visit websites on users' behalf to answer questions, integrate APIs, and complete tasks. Together with the ora.ai research lab, we observed 1,033 agent runs across 25 sites and ran 190 controlled fetch probes.\" \"This guide reflects data collected from June through August 2026. Ora continues to expand the datasets with more sites, models, harnesses, and probes. Future guide versions may update the figures and recommendations as the sample grows. Every recommendation depends on the agent reaching the page and finding the answer in the returned HTML.\" \"v1.0 · August 2026 · first public version.\""
    last_modified: 2026-09-03T00:00:00Z
  - id: ref-2
    resource: "https://vercel.com/kb/guide/make-your-site-readable-by-ai-agents.md"
    title: "Vercel and Ora research lab — Make your site readable by AI agents (Markdown twin, frontmatter published: 2026-08-28, last_updated: 2026-09-03): \"title: Make your site readable by AI agents\" \"published: 2026-08-28\" \"last_updated: 2026-09-03\" \"In that controlled experiment, two kinds of change prevented answer retrieval. One hid the answer behind JavaScript from a fetch-only client. The other returned 403 to both clients. Agents must find and read a site's content before they can act on it.\" \"In the controlled probes, adding llms.txt, JSON-LD, sitemaps, Markdown mirrors, and similar features did not change whether the clients could retrieve an answer already present in server-rendered HTML.\" \"Publish each answer on one canonical, current first-party docs page. In these runs, docs were the most common grounding source. Of the 844 grounded answers, 394, or 47%, traced to a docs page. Public sites do not need a robots.txt file to permit crawling.\" \"Depending on the file type, 86-97% of attributed fetches came through links rather than guessed paths.\" \"Across the study harnesses, 2,259 of 3,470 web fetches, or 65.1%, requested Markdown. When a fetch specified any format, it selected Markdown in 2,259 of 2,361 cases, or 95.7%.\" \"If you offer Markdown through content negotiation, keep the response metadata accurate. A request with Accept: text/markdown should receive Markdown with Content-Type: text/markdown; charset=utf-8. Include Vary: Accept so caches keep the HTML and Markdown responses separate. If Markdown is unavailable, return HTML with its correct content type.\" \"Use is-agentic.com to check domains against a wider set of tests. The Ora research lab runs and scores every scan. Expect the checks and their weights to shift as new agent behavior gets measured.\""
    last_modified: 2026-09-03T00:00:00Z
  - id: ref-3
    resource: "https://timothyjordan.com/blog/2026/03/23/agent-readability-spec.html"
    title: "Agent Readability: A Specification for AI-Optimized Websites — Timothy Jordan — author-hosted cross-post of the original Vercel KB specification (published 23 March 2026): \"Agent Readability: A Specification for AI-Optimized Websites Mar 23, 2026 Cross-posted from Vercel KB AI agents — tools like ChatGPT, Claude, Copilot, and Cursor — are becoming a primary way people discover and consume web content. Agent readability is a set of best practices that make your website accessible to AI agents and assistants.\" \"Discovery — Can agents find your pages? (llms.txt, sitemaps, robots.txt) Structure — Can agents parse your pages? (meta tags, headings, structured data, markdown mirrors) Context — Can agents understand your content? (skill files, content negotiation, code documentation)\" \"Use 3 or more headings (h1–h3) per page to create a clear structure. Well-structured pages produce better embeddings and allow agents to extract specific sections. Signal-to-noise ratio Maintain a text-to-HTML ratio above 15%.\" \"score = round((passed checks / total checks) x 100) Only checks with a pass status count toward the numerator. Checks that fail, warn, or error do not. The total is the sum of all site-wide checks plus all per-page checks across every discovered page.\""
    last_modified: 2026-03-23T00:00:00Z
  - id: ref-4
    resource: "https://raw.githubusercontent.com/ThatGuySam/agentready/main/examples/vercel/agent-readability-spec.sample.html.txt"
    title: "ThatGuySam/agentready — archived HTML snapshot of Vercel KB Agent Readability: A Specification for AI-Optimized Websites (last updated 23 March 2026): \"Agent Readability: A Specification for AI-Optimized Websites | Vercel Knowledge Base\" \"When an agent visits your site, it needs to quickly find, read, and understand your pages. Sites that are easy for agents to navigate get cited more often, surface in more answers, and reach a wider audience.\""
    last_modified: 2026-03-23T00:00:00Z
  - id: ref-5
    resource: "https://registry.npmjs.org/@vercel/agent-readability"
    title: "npm registry — @vercel/agent-readability package metadata and README (v0.1.0 published 30 March 2026, v0.7.0 published 4 September 2026): \"\"name\":\"@vercel/agent-readability\"\" \"25 weighted checks across 4 categories. Score 0-100.\" \"1. Known UA patterns. 30+ agents (ClaudeBot, GPTBot, Cursor, Perplexity, etc.)\" \"2. Signature-Agent header. ChatGPT agent via RFC 9421.\" \"3. sec-fetch-mode heuristic. Unknown bots lacking browser fingerprints.\""
    last_modified: 2026-09-04T00:00:00Z
  - id: ref-6
    resource: "https://raw.githubusercontent.com/timothyjordan/a14y/main/README.md"
    title: "Timothy Jordan — timothyjordan/a14y GitHub repository README: \"a14y (shorthand for agentreadability) is an open spec for making any website discoverable, parseable, and comprehensible to those agents, a versioned scorecard that operationalizes the spec, and a CLI plus Chrome extension that score any site against it.\" \"The CLI also ships under two alias names for discoverability: agentready and agentreadability. All three run the same binary.\""
    last_modified: 2026-05-08T00:00:00Z
  - id: ref-7
    resource: "https://a14y.dev/spec/"
    title: "a14y — The Agent Readability spec (v0.2.0 released 6 April 2026): \"v0.2.0 · released 2026-04-06 Agent readability, defined. An open set of practices that make any website discoverable, parseable, and comprehensible to AI agents. The scorecard operationalizes the spec into pass-or-fail checks.\""
    last_modified: 2026-04-06T00:00:00Z
  - id: ref-8
    resource: "https://a14y.dev/scorecards/"
    title: "Scorecards — a14y — Scorecards index and methodology (v0.2.0 flat-pool-v1, 38 pinned checks): \"Each scorecard pins exactly one algorithm. v0.2.0 ships flat-pool-v1; the v0.3.0 draft introduces per-check-mean-v1 to fix a page-count weighting issue in flat-pool.\" \"released 2026-04-06 · 38 checks pinned\""
    last_modified: 2026-04-06T00:00:00Z
  - id: ref-9
    resource: "https://a14y.dev/scorecards/0.2.0.md"
    title: "a14y — Scorecard v0.2.0 Markdown mirror (last_updated 6 October 2026): \"title: Scorecard v0.2.0 · a14y agent readability\" \"Initial public scorecard for the Agent Readability Spec. Covers site discoverability (llms.txt, robots.txt, sitemaps, AGENTS.md) and per-page HTML/markdown/code/API checks.\""
    last_modified: 2026-10-06T00:00:00Z
  - id: ref-10
    resource: "https://a14y.dev/.well-known/agent-skills/index.json"
    title: "a14y — Agent Skills Discovery index (/.well-known/agent-skills/index.json, updated 6 October 2026): \"\"name\": \"a14y\"\" \"\"url\": \"/.well-known/agent-skills/a14y/SKILL.md\"\" \"\"spec\": \"https://agentskills.io/specification\"\""
    last_modified: 2026-10-06T00:00:00Z
  - id: ref-11
    resource: "https://raw.githubusercontent.com/agentready-org/standard/main/README.md"
    title: "agentready-org/standard — AgentReady open standard repository README (written by ora.ai and Vercel): \"An open standard for the agentic web — a product implements AgentReady when AI agents can use it from discovery to completion.\" \"1,033 real agent runs across 25 public product sites (Stripe, Notion, Datadog, Twilio, and others).\""
    last_modified: 2026-10-29T00:00:00Z
  - id: ref-12
    resource: "https://raw.githubusercontent.com/agentready-org/standard/main/spec/spec.md"
    title: "agentready-org/standard — spec/spec.md (AgentReady — Make your site readable by AI agents, v2.0 October 2026): \"AgentReady — Make your site readable by AI agents\" \"Status: v2.0 — October 2026\" \"Normative weight is derived from measurement. In the studies, only two changes caused agents to fail outright: hiding answers behind JavaScript and blocking access. Those, the prohibitions that make failures indistinguishable from successes (dishonest status codes), and, for products agents are expected to act on, an interface to act through (AR-ACT-10) are MUST.\""
    last_modified: 2026-10-29T00:00:00Z
  - id: ref-13
    resource: "https://raw.githubusercontent.com/agentready-org/standard/main/CHANGELOG.md"
    title: "agentready-org/standard — CHANGELOG.md (v1.0.0 April 2026, v1.0 August 2026, v2.0 October 2026): \"v2.0 — October 2026\" \"New practice: AR-ACT-10 an MCP server or a CLI (MUST, at least one) for products agents are expected to act on.\" \"v1.0 — August 2026\" \"First public version of \"Make your site readable by AI agents.\"\""
    last_modified: 2026-10-29T00:00:00Z
  - id: ref-14
    resource: "https://raw.githubusercontent.com/agentready-org/standard/main/data/fetchability.csv"
    title: "agentready-org/standard — data/fetchability.csv (190 controlled fetch probes across 19 site configurations): \"returns 403 to agent user-agents, hostile configuration\" \"server-rendered page, no extras\""
    last_modified: 2026-08-28T00:00:00Z
  - id: ref-15
    resource: "https://www.agentready.org/"
    title: "The open standard for agent readiness — AgentReady — canonical web surface (www.agentready.org): \"AgentReady is an open specification of what a product should implement to be usable by AI agents, from discovery to completion. It does not promote a protocol or a vendor. The practices below are based on measured agent behavior, and are revised as that behavior changes.\""
    last_modified: 2026-10-29T00:00:00Z
  - id: ref-16
    resource: "https://www.agentready.org/llms.txt"
    title: "AgentReady — /llms.txt machine-readable summary (v2.0 October 2026): \"AgentReady is the open standard for agent readiness: a spec of the protocols, standards, and conventions a product should implement to be found, read, and used by AI agents, from discovery to completion.\" \"Developed in collaboration with Ora, Vercel, Stripe, Auth0, Mintlify, Tavily, Google, Elementor, and Netlify.\""
    last_modified: 2026-10-29T00:00:00Z
  - id: ref-17
    resource: "https://is-agentic.com/"
    title: "Ora research lab — Is Agentic: AI Agent Readiness Score for your Site and App public scanner and scoring methodology: \"Is Agentic scores the public parts of a website that AI agents can discover, retrieve, understand, and use. Essential checks carry most of the score and cover fundamentals such as server-rendered content, correct HTTP behavior, clear document structure, recoverable errors, usable controls, and the identity metadata agents rely on for citations and previews\""
    last_modified: 2026-10-29T00:00:00Z
  - id: ref-18
    resource: "https://is-agentic.com/api/v1/report?url=vercel.com"
    title: "Is Agentic API v1 — live JSON readiness report for vercel.com (scanned 28 October 2026): \"\"target\":\"https://vercel.com\"\" \"\"report_url\":\"https://is-agentic.com/scan/vercel.com\"\""
    last_modified: 2026-10-28T00:00:00Z
  - id: ref-19
    resource: "https://ora.ai/blog/what-agents-actually-reach"
    title: "What agents actually reach — Ora research — What agents actually reach field report (published 16 June 2026): \"Agents lean hard on the human web. They reach for homepages and docs - the same pages a person would, in the same order. The purpose-built, agent-first files get used far less. But where those files exist and get reached, they convert better than anything else. The bottleneck is discovery.\""
    last_modified: 2026-06-16T00:00:00Z
  - id: ref-20
    resource: "https://ora.ai/research"
    title: "Ora research — Research: how ready is the web for agents? ecosystem score distribution across scanned sites: \"The ecosystem average sits at 43 out of 100; half of all sites score below 42, half above. On the letter ladder, Grade D alone holds 37 % of all sites - the curve is narrow and concentrated.\""
    last_modified: 2026-10-29T00:00:00Z
  - id: ref-21
    resource: "https://vercel.com/docs/agent-resources.md"
    title: "Vercel Docs — Agent Resources Markdown twin (last_updated 15 September 2026): \"Every documentation page is available as Markdown, and site-wide indexes help agents find pages and understand how they connect.\""
    last_modified: 2026-09-15T00:00:00Z
  - id: ref-22
    resource: "https://vercel.com/docs/agent-resources/markdown-access.md"
    title: "Vercel Docs — Markdown and Agent Discovery Markdown twin (last_updated 17 September 2026): \"A canonical docs URL such as https://vercel.com/docs/functions returns HTML to an ordinary browser request. The same URL returns text/markdown when the request prefers Markdown, or when Vercel detects an AI agent without a format preference.\" \"Every Markdown response sets the Vary: Accept header.\" \"Vercel builds a nightly cross-link graph across Vercel docs, the Vercel Knowledge Base, nextjs.org, ai-sdk.dev, and other Vercel documentation sites.\""
    last_modified: 2026-09-17T00:00:00Z
  - id: ref-23
    resource: "https://vercel.com/docs/sitemap.md"
    title: "Vercel Docs — /docs/sitemap.md semantic documentation index: \"# Vercel Documentation Sitemap\" \"This file is a high-level semantic index of the documentation.\""
    last_modified: 2026-10-29T00:00:00Z
  - id: ref-24
    resource: "https://vercel.com/docs/agent-resources.graph.md"
    title: "Vercel Docs — /docs/agent-resources.graph.md per-page cross-link map: \"# Cross-link map: Agent Resources (/docs/agent-resources)\" \"Full graph as JSON: https://vercel.com/docs/graph.json\""
    last_modified: 2026-10-28T00:00:00Z
  - id: ref-25
    resource: "https://vercel.com/blog/making-agent-friendly-pages-with-content-negotiation"
    title: "Making agent-friendly pages with content negotiation — Zach Cowan and Mitul Shah — Vercel Field Engineering blog post (published 3 February 2026): \"Making agent-friendly pages with content negotiation\" \"What agents need is a way to request just the text content of a page, without the browser-specific markup. Content negotiation solves this. It's a standard HTTP mechanism where the client specifies its preferred format via the Accept header, and the server returns the matching representation.\""
    last_modified: 2026-02-03T00:00:00Z
  - id: ref-26
    resource: "https://vercel.com/kb/guide/how-to-serve-documentation-for-agents.md"
    title: "Anthony Shew — How to serve documentation for agents Markdown twin (published 14 January 2026, updated 17 January 2026): \"title: How to serve documentation for agents\" \"description: Learn how to serve markdown to agents and HTML for humans from the same URL\""
    last_modified: 2026-01-17T00:00:00Z
  - id: ref-27
    resource: "https://vercel.com/changelog/docs-pages-support-markdown-responses"
    title: "Docs pages support Markdown responses — Anthony Shew — Vercel Changelog entry (published 13 January 2026): \"Docs pages support Markdown responses\" \"You can now request Vercel documentation as Markdown by sending the Accept header with the value text/markdown.\""
    last_modified: 2026-01-13T00:00:00Z
  - id: ref-28
    resource: "https://vercel.com/kb/guide/agentic-commerce-readiness.md"
    title: "Vercel Knowledge Base — Making your catalog discoverable to AI agents Markdown twin (published 1 October 2026): \"AI shopping agents read your product pages before any JavaScript runs. The crawlers behind them fetch your HTML and never execute it, so a storefront that assembles its catalog in the browser returns a shell with no product name, price, or availability in it.\""
    last_modified: 2026-10-01T00:00:00Z
  - id: ref-29
    resource: "https://vercel.com/docs/cli/agent.md"
    title: "Vercel CLI Docs — vercel agent command Markdown twin (last_updated 6 June 2026): \"The vercel agent command writes a section related to Vercel deployment best practices in your project's agent guidance file so coding agents and assistants have context for how the project should build, deploy, and integrate with Vercel features.\""
    last_modified: 2026-06-06T00:00:00Z
  - id: ref-30
    resource: "https://raw.githubusercontent.com/vercel/vercel-plugin/main/README.md"
    title: "Vercel — vercel/vercel-plugin GitHub README (is-agentic skill and 37 bundled skills): \"Score how ready a website, domain, or public MCP endpoint is for AI agents, and act on the Is Agentic report\""
    last_modified: 2026-10-29T00:00:00Z
  - id: ref-31
    resource: "https://nextjs.org/docs/app/guides/ai-agents.md"
    title: "Next.js Docs — How Next.js supports AI coding agents Markdown twin (lastUpdated 15 October 2026): \"Next.js 16.2 and later include documentation that matches the installed framework version.\" \"The bundled copy gives coding agents access to the APIs and patterns supported by the project without a network request or external lookup.\""
    last_modified: 2026-10-15T00:00:00Z
  - id: ref-32
    resource: "https://dacharycarey.com/2026/03/28/how-to-evaluate-platform-written-spec/"
    title: "How to Evaluate a Platform-Written Spec — Dachary Carey — independent technical critique of Vercel Agent Readability Spec (published 28 March 2026): \"Recently, Vercel published an Agent Readability Spec (Timothy Jordan, last updated March 23, no publication date), a scoring system for how well websites serve AI agents. It covers llms.txt, markdown mirrors, content negotiation, structured data, robots.txt, and a checklist of other recommendations. Four days later, they followed up with \"Make Your Documentation Readable by AI Agents\" (Rich Haines; March 27), a companion implementation guide that references the spec article and adds its own recommendations.\" \"My spec explicitly scopes to real-time coding agents (Claude Code, Cursor, Copilot) and excludes training crawlers (GPTBot, ClaudeBot). Both Vercel articles conflate them.\""
    last_modified: 2026-03-28T00:00:00Z
  - id: ref-33
    resource: "https://agentdocsspec.com/"
    title: "Dachary Carey — Agent-Friendly Documentation Spec landing page (agentdocsspec.com): \"Documentation sites are increasingly consumed by coding agents (Claude Code, Cursor, GitHub Copilot, and others) rather than human readers. But most sites weren't built for this access pattern. Agents hit truncation limits, get walls of CSS instead of content, can't follow cross-host redirects, and don't know about emerging discovery mechanisms like llms.txt.\""
    last_modified: 2026-09-13T00:00:00Z
  - id: ref-34
    resource: "https://agentdocsspec.com/spec/web/index.md"
    title: "Dachary Carey + community contributors — Web Documentation Delivery Spec v0.6.0 (dated 13 September 2026): \"| Version | 0.6.0 |\" \"| Date | 2026-09-13 |\" \"| Author | Dachary Carey + community contributors |\""
    last_modified: 2026-09-13T00:00:00Z
  - id: ref-35
    resource: "https://raw.githubusercontent.com/agent-ecosystem/afdocs/main/README.md"
    title: "agent-ecosystem/afdocs — AFDocs CLI reference implementation README: \"Test your documentation site against the Agent-Friendly Documentation Spec. AFDocs runs checks across 7 categories to measure how well AI coding agents can discover, navigate, and consume your docs. Powering Agent Score by Fern.\" \"Implements spec v0.6.0 (2026-09-13).\""
    last_modified: 2026-09-13T00:00:00Z
  - id: ref-36
    resource: "https://buildwithfern.com/post/agent-score"
    title: "How we built Agent Score to improve Fern Docs — Kapil Gowru (Fern) — How we built Agent Score to improve Fern Docs (published 7 April 2026): \"How we built Agent Score to improve Fern Docs\" \"I partnered with Dachary Carey to build Agent Score, a free, open-source benchmark that grades your docs 0–100 on agent-readiness.\""
    last_modified: 2026-04-07T00:00:00Z
  - id: ref-37
    resource: "https://docs.readme.com/main/docs/ai-discoverability"
    title: "Discoverability — ReadMe Documentation — AI Discoverability settings and Agent-Friendly Docs spec scoring: \"Discoverability controls how easily AI agents and assistants can find, crawl, and understand your documentation. ReadMe crawls a sample of your published pages and scores your project against the Agent-Friendly Docs spec, so you can see at a glance where your docs stand and which switches to flip to improve the score. Every setting on this page has a sensible default.\""
    last_modified: 2026-10-29T00:00:00Z
  - id: ref-38
    resource: "https://rebilder.com/compare/vercel-agent-readability"
    title: "Rebilder and @vercel/agent-readability — Rebilder — technical comparison of @vercel/agent-readability and Rebilder ARS 0.3.0 (checked 28 September 2026): \"An MIT-licensed npm package with framework middleware and an audit CLI. The latest version on September 28, 2026 was 0.7.0.\" \"Vercel's Agent Readability Spec, a guide on vercel.com now titled \"Make your site readable by AI agents\". The Rebilder Agent Readability Spec (ARS) 0.3.0, published at rebilder.com/spec/ars with a conformance corpus. The two specs share a name and are separate documents.\""
    last_modified: 2026-09-28T00:00:00Z
  - id: ref-39
    resource: "https://dev.to/earlgreyhot1701d/i-built-it-twice-didnt-place-twice-then-google-and-vercel-shipped-it-ldd"
    title: "I built it twice. Didn't place twice. Then Google and Vercel shipped it. — Earl Grey — DEV Community commentary on vendor-published agent readability specifications (published 4 September 2026): \"A hosting company wrote a standard and the ecosystem picked it up. No standards body was in the room. That is the adoption split again, one level higher. llms.txt asks whether you added a file. The spec asks whether you followed a document a vendor published.\""
    last_modified: 2026-09-04T00:00:00Z
  - id: ref-40
    resource: "https://odinmind.com/the-agentic-web-ai-agents-shall-inherit-the-earth/"
    title: "The Agentic Web: AI Agents shall inherit the earth — Odin Mind — editorial analysis citing Vercel Agent Readability Specification (published 5 April 2026): \"Vercel's Agent Readability Specification gives developers a concrete checklist: serve llms.txt files, publish both XML and Markdown sitemaps, annotate HTML properly.\""
    last_modified: 2026-04-05T00:00:00Z
  - id: ref-41
    resource: "https://aihumanlove.com/courses/ai-seo-foundations/lesson-03.html"
    title: "How AI Agents Read a Website — AIHumanLove — AI SEO Foundations Lesson 3 (updated 9 October 2026): \"Vercel — Agent Readability Specification\""
    last_modified: 2026-10-09T00:00:00Z
  - id: ref-42
    resource: "https://api.github.com/repos/GoogleChrome/lighthouse/releases/tags/v13.3.0"
    title: "GoogleChrome/lighthouse — Release v13.3.0 adding Agentic Browsing category (published 7 May 2026): \"New agentic browsing category added to default config\""
    last_modified: 2026-05-07T00:00:00Z
  - id: ref-43
    resource: "https://api.github.com/repos/GoogleChrome/lighthouse/releases/tags/v13.5.0"
    title: "GoogleChrome/lighthouse — Release v13.5.0 adding ard-schema audit and agent discovery grouping (published 18 September 2026): \"ard-schema: add Agent Resource Discovery audit\" \"config: group llms.txt and ARD audits under agent discovery\""
    last_modified: 2026-09-18T00:00:00Z
  - id: ref-44
    resource: "https://developer.chrome.com/docs/lighthouse/agentic-browsing/scoring"
    title: "Chrome for Developers — Agentic Browsing scoring Lighthouse documentation: \"Unlike other Lighthouse categories, the Agentic Browsing category does not have a weighted average score from 0 to 100. Because the standards for the agentic web are still emerging, the current focus is to gather data and provide actionable signals rather than a definitive ranking.\" \"A fractional score: A ratio showing how many agentic readiness checks your site passes.\""
    last_modified: 2026-10-29T00:00:00Z
  - id: ref-45
    resource: "https://developer.chrome.com/docs/lighthouse/agentic-browsing/llms-txt"
    title: "Chrome for Developers — Page has a valid llms.txt file Lighthouse audit documentation: \"Lighthouse flags the pages if a server error occurs when attempting to retrieve the llms.txt file. If the file is not provided by the server (resulting in a 404), the audit is marked as Not Applicable (N/A), as providing the file is optional at the moment.\""
    last_modified: 2026-10-29T00:00:00Z
---

# Agent Readability Spec

Agent Readability Spec (ARS) is an agentic-web standard (document format) at step 4, Read, governed by Original checklist: Timothy Jordan at Vercel; its status is "Original KB URL returns HTTP 301 / 308 Permanent Redirect to Make your site readable by AI agents | Vercel Knowledge Base". It answers: "How readable and actionable is this site for AI agents from discovery to completion?" Implementation status is shipping (@vercel/agent-readability, a14y, AgentReady v2.0, Ora is-agentic.com scanner, Vercel documentation estate), last verified 2026-10-29.

# Overview

The **Agent Readability Spec** originated in March 2026 as a Vercel Knowledge Base specification authored by Timothy Jordan (*Agent Readability: A Specification for AI-Optimized Websites | Timothy Jordan*, cross-posted from *Agent Readability: A Specification for AI-Optimized Websites | Vercel Knowledge Base*) that defined website practices for AI coding assistants and web agents across discovery, page structure, and contextual metadata, paired with an unweighted pass-fraction scoring formula [^ref-3][^ref-4]. Rather than introducing a new wire protocol or bespoke schema format, the March 2026 checklist assembled existing site-level and page-level conventions—including root `/llms.txt` indexes, `robots.txt` bot rules, XML and Markdown sitemaps (`/sitemap.xml` and `/sitemap.md`), repository or site `AGENTS.md` instruction files, canonical HTML link elements, Schema.org JSON-LD blocks, section heading hierarchy, text-to-HTML density thresholds, fenced language-tagged code blocks, OpenAPI reference links, `.md` page mirrors, `<link rel="alternate" type="text/markdown">` tags, RFC 8288 `Link: <...>; rel="canonical"` response headers, and `Accept: text/markdown` HTTP content negotiation—into a single audit rubric [^ref-3][^ref-4].

In late March 2026, Vercel shipped the companion MIT-licensed npm package `@vercel/agent-readability`, combining a site-audit command-line interface with routing middleware for Next.js, SvelteKit, Nuxt, and Vercel Edge deployments that detects agent requests through user-agent strings, cryptographic signature headers, and browser-header heuristics and rewrites matching requests to Markdown endpoints with `Vary: Accept` cache separation [^ref-5][^ref-38]. In April and May 2026, Jordan also extracted the checklist into the standalone open-source `a14y` project (*The Agent Readability spec · a14y* and *Scorecards · a14y*), pinning versioned check manifests so historical site audits remain reproducible when scoring algorithms evolve [^ref-6][^ref-7][^ref-8][^ref-9].

In August 2026, Vercel and the Ora research lab replaced the original Knowledge Base checklist and configured `https://vercel.com/kb/guide/agent-readability-spec` (and its `.md` twin) to issue a permanent redirect to *Make your site readable by AI agents | Vercel Knowledge Base* (`https://vercel.com/kb/guide/make-your-site-readable-by-ai-agents`) [^ref-1][^ref-2]. Grounded in turn-by-turn agent traces across public product sites and a controlled single-variable fetchability experiment published in the `agentready-org/standard` repository (*AgentReady // The open standard for agent readiness*), the August 2026 guide and its October 2026 successor specification (`AgentReady`) substantially revised the assumptions of the earlier checklist [^ref-1][^ref-2][^ref-11][^ref-12][^ref-15]. Because controlled testing showed that answer extraction failed outright only when content was hidden behind client-side JavaScript or blocked at the HTTP layer—whereas adding `llms.txt`, sitemaps, JSON-LD, or Markdown mirrors did not alter retrieval when the answer already appeared in server-rendered HTML—the revised specification reorganizes requirements into three sequential stages (`find`, `read`, and `act`) and reserves mandatory compliance levels for unblocked access, server-rendered HTML content, honest HTTP status codes, and an actionable programmatic interface (an MCP server or CLI) for products that agents are expected to operate [^ref-2][^ref-12][^ref-13][^ref-14].

# Examples

```http
GET {origin}/llms.txt, /robots.txt, /sitemap.xml, /sitemap.md, /AGENTS.md, Accept: text/markdown
# parse: Evaluated per surface across site-level and page-level checks by @vercel/agent-readability audit, a14y, or is-agentic.com
# on absent: A site failing most checks is not in violation of anything; the spec is a voluntary readiness checklist
```

# Knowledge Graph

* **Lifecycle Step**: [4 Read](/category/read/index.md)
* **Pairs with**: [llms.txt](/okf/protocols/llms-txt.md), [AGENTS.md](/okf/protocols/agents-md.md), [Markdown for Agents](/okf/protocols/cloudflare-markdown-for-agents.md), [AI crawler tokens in robots.txt](/okf/protocols/ai-crawler-tokens-robots.md), [api-catalog (RFC 9727)](/okf/protocols/api-catalog-rfc9727.md), [AI Catalog well-known file](/okf/protocols/ai-catalog-well-known-file.md)
* **Compared in**: [HTML DOM & RSS vs Accept: text/markdown & OKF (classic vs agentic)](/compare/html-dom-vs-markdown-for-agents/index.md), [Agent Readability Spec vs llms.txt](/compare/vercel-agent-readability-vs-llms-txt/index.md), [Open Knowledge Format vs Agent Readability Spec](/compare/okf-vs-vercel-agent-readability/index.md), [Markdown for Agents vs Agent Readability Spec](/compare/cloudflare-markdown-for-agents-vs-vercel-agent-readability/index.md), [Agent Readability Spec vs AGENTS.md](/compare/vercel-agent-readability-vs-agents-md/index.md)

[^ref-1]: Make your site readable by AI agents — Vercel and Ora research lab — Make your site readable by AI agents (HTML guide, published 28 August 2026, updated 3 September 2026): "Make your site readable by AI agents" "AI agents visit websites on users' behalf to answer questions, integrate APIs, and complete tasks. Together with the ora.ai research lab, we observed 1,033 agent runs across 25 sites and ran 190 controlled fetch probes." "This guide reflects data collected from June through August 2026. Ora continues to expand the datasets with more sites, models, harnesses, and probes. Future guide versions may update the figures and recommendations as the sample grows. Every recommendation depends on the agent reaching the page and finding the answer in the returned HTML." "v1.0 · August 2026 · first public version." (https://vercel.com/kb/guide/make-your-site-readable-by-ai-agents)
[^ref-2]: Vercel and Ora research lab — Make your site readable by AI agents (Markdown twin, frontmatter published: 2026-08-28, last_updated: 2026-09-03): "title: Make your site readable by AI agents" "published: 2026-08-28" "last_updated: 2026-09-03" "In that controlled experiment, two kinds of change prevented answer retrieval. One hid the answer behind JavaScript from a fetch-only client. The other returned 403 to both clients. Agents must find and read a site's content before they can act on it." "In the controlled probes, adding llms.txt, JSON-LD, sitemaps, Markdown mirrors, and similar features did not change whether the clients could retrieve an answer already present in server-rendered HTML." "Publish each answer on one canonical, current first-party docs page. In these runs, docs were the most common grounding source. Of the 844 grounded answers, 394, or 47%, traced to a docs page. Public sites do not need a robots.txt file to permit crawling." "Depending on the file type, 86-97% of attributed fetches came through links rather than guessed paths." "Across the study harnesses, 2,259 of 3,470 web fetches, or 65.1%, requested Markdown. When a fetch specified any format, it selected Markdown in 2,259 of 2,361 cases, or 95.7%." "If you offer Markdown through content negotiation, keep the response metadata accurate. A request with Accept: text/markdown should receive Markdown with Content-Type: text/markdown; charset=utf-8. Include Vary: Accept so caches keep the HTML and Markdown responses separate. If Markdown is unavailable, return HTML with its correct content type." "Use is-agentic.com to check domains against a wider set of tests. The Ora research lab runs and scores every scan. Expect the checks and their weights to shift as new agent behavior gets measured." (https://vercel.com/kb/guide/make-your-site-readable-by-ai-agents.md)
[^ref-3]: Agent Readability: A Specification for AI-Optimized Websites — Timothy Jordan — author-hosted cross-post of the original Vercel KB specification (published 23 March 2026): "Agent Readability: A Specification for AI-Optimized Websites Mar 23, 2026 Cross-posted from Vercel KB AI agents — tools like ChatGPT, Claude, Copilot, and Cursor — are becoming a primary way people discover and consume web content. Agent readability is a set of best practices that make your website accessible to AI agents and assistants." "Discovery — Can agents find your pages? (llms.txt, sitemaps, robots.txt) Structure — Can agents parse your pages? (meta tags, headings, structured data, markdown mirrors) Context — Can agents understand your content? (skill files, content negotiation, code documentation)" "Use 3 or more headings (h1–h3) per page to create a clear structure. Well-structured pages produce better embeddings and allow agents to extract specific sections. Signal-to-noise ratio Maintain a text-to-HTML ratio above 15%." "score = round((passed checks / total checks) x 100) Only checks with a pass status count toward the numerator. Checks that fail, warn, or error do not. The total is the sum of all site-wide checks plus all per-page checks across every discovered page." (https://timothyjordan.com/blog/2026/03/23/agent-readability-spec.html)
[^ref-4]: ThatGuySam/agentready — archived HTML snapshot of Vercel KB Agent Readability: A Specification for AI-Optimized Websites (last updated 23 March 2026): "Agent Readability: A Specification for AI-Optimized Websites | Vercel Knowledge Base" "When an agent visits your site, it needs to quickly find, read, and understand your pages. Sites that are easy for agents to navigate get cited more often, surface in more answers, and reach a wider audience." (https://raw.githubusercontent.com/ThatGuySam/agentready/main/examples/vercel/agent-readability-spec.sample.html.txt)
[^ref-5]: npm registry — @vercel/agent-readability package metadata and README (v0.1.0 published 30 March 2026, v0.7.0 published 4 September 2026): ""name":"@vercel/agent-readability"" "25 weighted checks across 4 categories. Score 0-100." "1. Known UA patterns. 30+ agents (ClaudeBot, GPTBot, Cursor, Perplexity, etc.)" "2. Signature-Agent header. ChatGPT agent via RFC 9421." "3. sec-fetch-mode heuristic. Unknown bots lacking browser fingerprints." (https://registry.npmjs.org/@vercel/agent-readability)
[^ref-6]: Timothy Jordan — timothyjordan/a14y GitHub repository README: "a14y (shorthand for agentreadability) is an open spec for making any website discoverable, parseable, and comprehensible to those agents, a versioned scorecard that operationalizes the spec, and a CLI plus Chrome extension that score any site against it." "The CLI also ships under two alias names for discoverability: agentready and agentreadability. All three run the same binary." (https://raw.githubusercontent.com/timothyjordan/a14y/main/README.md)
[^ref-7]: a14y — The Agent Readability spec (v0.2.0 released 6 April 2026): "v0.2.0 · released 2026-04-06 Agent readability, defined. An open set of practices that make any website discoverable, parseable, and comprehensible to AI agents. The scorecard operationalizes the spec into pass-or-fail checks." (https://a14y.dev/spec/)
[^ref-8]: Scorecards — a14y — Scorecards index and methodology (v0.2.0 flat-pool-v1, 38 pinned checks): "Each scorecard pins exactly one algorithm. v0.2.0 ships flat-pool-v1; the v0.3.0 draft introduces per-check-mean-v1 to fix a page-count weighting issue in flat-pool." "released 2026-04-06 · 38 checks pinned" (https://a14y.dev/scorecards/)
[^ref-9]: a14y — Scorecard v0.2.0 Markdown mirror (last_updated 6 October 2026): "title: Scorecard v0.2.0 · a14y agent readability" "Initial public scorecard for the Agent Readability Spec. Covers site discoverability (llms.txt, robots.txt, sitemaps, AGENTS.md) and per-page HTML/markdown/code/API checks." (https://a14y.dev/scorecards/0.2.0.md)
[^ref-10]: a14y — Agent Skills Discovery index (/.well-known/agent-skills/index.json, updated 6 October 2026): ""name": "a14y"" ""url": "/.well-known/agent-skills/a14y/SKILL.md"" ""spec": "https://agentskills.io/specification"" (https://a14y.dev/.well-known/agent-skills/index.json)
[^ref-11]: agentready-org/standard — AgentReady open standard repository README (written by ora.ai and Vercel): "An open standard for the agentic web — a product implements AgentReady when AI agents can use it from discovery to completion." "1,033 real agent runs across 25 public product sites (Stripe, Notion, Datadog, Twilio, and others)." (https://raw.githubusercontent.com/agentready-org/standard/main/README.md)
[^ref-12]: agentready-org/standard — spec/spec.md (AgentReady — Make your site readable by AI agents, v2.0 October 2026): "AgentReady — Make your site readable by AI agents" "Status: v2.0 — October 2026" "Normative weight is derived from measurement. In the studies, only two changes caused agents to fail outright: hiding answers behind JavaScript and blocking access. Those, the prohibitions that make failures indistinguishable from successes (dishonest status codes), and, for products agents are expected to act on, an interface to act through (AR-ACT-10) are MUST." (https://raw.githubusercontent.com/agentready-org/standard/main/spec/spec.md)
[^ref-13]: agentready-org/standard — CHANGELOG.md (v1.0.0 April 2026, v1.0 August 2026, v2.0 October 2026): "v2.0 — October 2026" "New practice: AR-ACT-10 an MCP server or a CLI (MUST, at least one) for products agents are expected to act on." "v1.0 — August 2026" "First public version of "Make your site readable by AI agents."" (https://raw.githubusercontent.com/agentready-org/standard/main/CHANGELOG.md)
[^ref-14]: agentready-org/standard — data/fetchability.csv (190 controlled fetch probes across 19 site configurations): "returns 403 to agent user-agents, hostile configuration" "server-rendered page, no extras" (https://raw.githubusercontent.com/agentready-org/standard/main/data/fetchability.csv)
[^ref-15]: The open standard for agent readiness — AgentReady — canonical web surface (www.agentready.org): "AgentReady is an open specification of what a product should implement to be usable by AI agents, from discovery to completion. It does not promote a protocol or a vendor. The practices below are based on measured agent behavior, and are revised as that behavior changes." (https://www.agentready.org/)
[^ref-16]: AgentReady — /llms.txt machine-readable summary (v2.0 October 2026): "AgentReady is the open standard for agent readiness: a spec of the protocols, standards, and conventions a product should implement to be found, read, and used by AI agents, from discovery to completion." "Developed in collaboration with Ora, Vercel, Stripe, Auth0, Mintlify, Tavily, Google, Elementor, and Netlify." (https://www.agentready.org/llms.txt)
[^ref-17]: Ora research lab — Is Agentic: AI Agent Readiness Score for your Site and App public scanner and scoring methodology: "Is Agentic scores the public parts of a website that AI agents can discover, retrieve, understand, and use. Essential checks carry most of the score and cover fundamentals such as server-rendered content, correct HTTP behavior, clear document structure, recoverable errors, usable controls, and the identity metadata agents rely on for citations and previews" (https://is-agentic.com/)
[^ref-18]: Is Agentic API v1 — live JSON readiness report for vercel.com (scanned 28 October 2026): ""target":"https://vercel.com"" ""report_url":"https://is-agentic.com/scan/vercel.com"" (https://is-agentic.com/api/v1/report?url=vercel.com)
[^ref-19]: What agents actually reach — Ora research — What agents actually reach field report (published 16 June 2026): "Agents lean hard on the human web. They reach for homepages and docs - the same pages a person would, in the same order. The purpose-built, agent-first files get used far less. But where those files exist and get reached, they convert better than anything else. The bottleneck is discovery." (https://ora.ai/blog/what-agents-actually-reach)
[^ref-20]: Ora research — Research: how ready is the web for agents? ecosystem score distribution across scanned sites: "The ecosystem average sits at 43 out of 100; half of all sites score below 42, half above. On the letter ladder, Grade D alone holds 37 % of all sites - the curve is narrow and concentrated." (https://ora.ai/research)
[^ref-21]: Vercel Docs — Agent Resources Markdown twin (last_updated 15 September 2026): "Every documentation page is available as Markdown, and site-wide indexes help agents find pages and understand how they connect." (https://vercel.com/docs/agent-resources.md)
[^ref-22]: Vercel Docs — Markdown and Agent Discovery Markdown twin (last_updated 17 September 2026): "A canonical docs URL such as https://vercel.com/docs/functions returns HTML to an ordinary browser request. The same URL returns text/markdown when the request prefers Markdown, or when Vercel detects an AI agent without a format preference." "Every Markdown response sets the Vary: Accept header." "Vercel builds a nightly cross-link graph across Vercel docs, the Vercel Knowledge Base, nextjs.org, ai-sdk.dev, and other Vercel documentation sites." (https://vercel.com/docs/agent-resources/markdown-access.md)
[^ref-23]: Vercel Docs — /docs/sitemap.md semantic documentation index: "# Vercel Documentation Sitemap" "This file is a high-level semantic index of the documentation." (https://vercel.com/docs/sitemap.md)
[^ref-24]: Vercel Docs — /docs/agent-resources.graph.md per-page cross-link map: "# Cross-link map: Agent Resources (/docs/agent-resources)" "Full graph as JSON: https://vercel.com/docs/graph.json" (https://vercel.com/docs/agent-resources.graph.md)
[^ref-25]: Making agent-friendly pages with content negotiation — Zach Cowan and Mitul Shah — Vercel Field Engineering blog post (published 3 February 2026): "Making agent-friendly pages with content negotiation" "What agents need is a way to request just the text content of a page, without the browser-specific markup. Content negotiation solves this. It's a standard HTTP mechanism where the client specifies its preferred format via the Accept header, and the server returns the matching representation." (https://vercel.com/blog/making-agent-friendly-pages-with-content-negotiation)
[^ref-26]: Anthony Shew — How to serve documentation for agents Markdown twin (published 14 January 2026, updated 17 January 2026): "title: How to serve documentation for agents" "description: Learn how to serve markdown to agents and HTML for humans from the same URL" (https://vercel.com/kb/guide/how-to-serve-documentation-for-agents.md)
[^ref-27]: Docs pages support Markdown responses — Anthony Shew — Vercel Changelog entry (published 13 January 2026): "Docs pages support Markdown responses" "You can now request Vercel documentation as Markdown by sending the Accept header with the value text/markdown." (https://vercel.com/changelog/docs-pages-support-markdown-responses)
[^ref-28]: Vercel Knowledge Base — Making your catalog discoverable to AI agents Markdown twin (published 1 October 2026): "AI shopping agents read your product pages before any JavaScript runs. The crawlers behind them fetch your HTML and never execute it, so a storefront that assembles its catalog in the browser returns a shell with no product name, price, or availability in it." (https://vercel.com/kb/guide/agentic-commerce-readiness.md)
[^ref-29]: Vercel CLI Docs — vercel agent command Markdown twin (last_updated 6 June 2026): "The vercel agent command writes a section related to Vercel deployment best practices in your project's agent guidance file so coding agents and assistants have context for how the project should build, deploy, and integrate with Vercel features." (https://vercel.com/docs/cli/agent.md)
[^ref-30]: Vercel — vercel/vercel-plugin GitHub README (is-agentic skill and 37 bundled skills): "Score how ready a website, domain, or public MCP endpoint is for AI agents, and act on the Is Agentic report" (https://raw.githubusercontent.com/vercel/vercel-plugin/main/README.md)
[^ref-31]: Next.js Docs — How Next.js supports AI coding agents Markdown twin (lastUpdated 15 October 2026): "Next.js 16.2 and later include documentation that matches the installed framework version." "The bundled copy gives coding agents access to the APIs and patterns supported by the project without a network request or external lookup." (https://nextjs.org/docs/app/guides/ai-agents.md)
[^ref-32]: How to Evaluate a Platform-Written Spec — Dachary Carey — independent technical critique of Vercel Agent Readability Spec (published 28 March 2026): "Recently, Vercel published an Agent Readability Spec (Timothy Jordan, last updated March 23, no publication date), a scoring system for how well websites serve AI agents. It covers llms.txt, markdown mirrors, content negotiation, structured data, robots.txt, and a checklist of other recommendations. Four days later, they followed up with "Make Your Documentation Readable by AI Agents" (Rich Haines; March 27), a companion implementation guide that references the spec article and adds its own recommendations." "My spec explicitly scopes to real-time coding agents (Claude Code, Cursor, Copilot) and excludes training crawlers (GPTBot, ClaudeBot). Both Vercel articles conflate them." (https://dacharycarey.com/2026/03/28/how-to-evaluate-platform-written-spec/)
[^ref-33]: Dachary Carey — Agent-Friendly Documentation Spec landing page (agentdocsspec.com): "Documentation sites are increasingly consumed by coding agents (Claude Code, Cursor, GitHub Copilot, and others) rather than human readers. But most sites weren't built for this access pattern. Agents hit truncation limits, get walls of CSS instead of content, can't follow cross-host redirects, and don't know about emerging discovery mechanisms like llms.txt." (https://agentdocsspec.com/)
[^ref-34]: Dachary Carey + community contributors — Web Documentation Delivery Spec v0.6.0 (dated 13 September 2026): "| Version | 0.6.0 |" "| Date | 2026-09-13 |" "| Author | Dachary Carey + community contributors |" (https://agentdocsspec.com/spec/web/index.md)
[^ref-35]: agent-ecosystem/afdocs — AFDocs CLI reference implementation README: "Test your documentation site against the Agent-Friendly Documentation Spec. AFDocs runs checks across 7 categories to measure how well AI coding agents can discover, navigate, and consume your docs. Powering Agent Score by Fern." "Implements spec v0.6.0 (2026-09-13)." (https://raw.githubusercontent.com/agent-ecosystem/afdocs/main/README.md)
[^ref-36]: How we built Agent Score to improve Fern Docs — Kapil Gowru (Fern) — How we built Agent Score to improve Fern Docs (published 7 April 2026): "How we built Agent Score to improve Fern Docs" "I partnered with Dachary Carey to build Agent Score, a free, open-source benchmark that grades your docs 0–100 on agent-readiness." (https://buildwithfern.com/post/agent-score)
[^ref-37]: Discoverability — ReadMe Documentation — AI Discoverability settings and Agent-Friendly Docs spec scoring: "Discoverability controls how easily AI agents and assistants can find, crawl, and understand your documentation. ReadMe crawls a sample of your published pages and scores your project against the Agent-Friendly Docs spec, so you can see at a glance where your docs stand and which switches to flip to improve the score. Every setting on this page has a sensible default." (https://docs.readme.com/main/docs/ai-discoverability)
[^ref-38]: Rebilder and @vercel/agent-readability — Rebilder — technical comparison of @vercel/agent-readability and Rebilder ARS 0.3.0 (checked 28 September 2026): "An MIT-licensed npm package with framework middleware and an audit CLI. The latest version on September 28, 2026 was 0.7.0." "Vercel's Agent Readability Spec, a guide on vercel.com now titled "Make your site readable by AI agents". The Rebilder Agent Readability Spec (ARS) 0.3.0, published at rebilder.com/spec/ars with a conformance corpus. The two specs share a name and are separate documents." (https://rebilder.com/compare/vercel-agent-readability)
[^ref-39]: I built it twice. Didn't place twice. Then Google and Vercel shipped it. — Earl Grey — DEV Community commentary on vendor-published agent readability specifications (published 4 September 2026): "A hosting company wrote a standard and the ecosystem picked it up. No standards body was in the room. That is the adoption split again, one level higher. llms.txt asks whether you added a file. The spec asks whether you followed a document a vendor published." (https://dev.to/earlgreyhot1701d/i-built-it-twice-didnt-place-twice-then-google-and-vercel-shipped-it-ldd)
[^ref-40]: The Agentic Web: AI Agents shall inherit the earth — Odin Mind — editorial analysis citing Vercel Agent Readability Specification (published 5 April 2026): "Vercel's Agent Readability Specification gives developers a concrete checklist: serve llms.txt files, publish both XML and Markdown sitemaps, annotate HTML properly." (https://odinmind.com/the-agentic-web-ai-agents-shall-inherit-the-earth/)
[^ref-41]: How AI Agents Read a Website — AIHumanLove — AI SEO Foundations Lesson 3 (updated 9 October 2026): "Vercel — Agent Readability Specification" (https://aihumanlove.com/courses/ai-seo-foundations/lesson-03.html)
[^ref-42]: GoogleChrome/lighthouse — Release v13.3.0 adding Agentic Browsing category (published 7 May 2026): "New agentic browsing category added to default config" (https://api.github.com/repos/GoogleChrome/lighthouse/releases/tags/v13.3.0)
[^ref-43]: GoogleChrome/lighthouse — Release v13.5.0 adding ard-schema audit and agent discovery grouping (published 18 September 2026): "ard-schema: add Agent Resource Discovery audit" "config: group llms.txt and ARD audits under agent discovery" (https://api.github.com/repos/GoogleChrome/lighthouse/releases/tags/v13.5.0)
[^ref-44]: Chrome for Developers — Agentic Browsing scoring Lighthouse documentation: "Unlike other Lighthouse categories, the Agentic Browsing category does not have a weighted average score from 0 to 100. Because the standards for the agentic web are still emerging, the current focus is to gather data and provide actionable signals rather than a definitive ranking." "A fractional score: A ratio showing how many agentic readiness checks your site passes." (https://developer.chrome.com/docs/lighthouse/agentic-browsing/scoring)
[^ref-45]: Chrome for Developers — Page has a valid llms.txt file Lighthouse audit documentation: "Lighthouse flags the pages if a server error occurs when attempting to retrieve the llms.txt file. If the file is not provided by the server (resulting in a 404), the audit is marked as Not Applicable (N/A), as providing the file is optional at the moment." (https://developer.chrome.com/docs/lighthouse/agentic-browsing/llms-txt)