answerrail (@answerrail/next, test-ai-ready & ai-crawler-check)

Content negotiation for Next.js & any web stack — serve clean Markdown to AI crawlers (PerplexityBot, ClaudeBot, OAI-SearchBot) via Accept: text/markdown while serving normal HTML to human browsers from a single URL.

1. Quick Start: Next.js Markdown Content Negotiation (Accept: text/markdown)

Install answerrail in your Next.js application:

npm install answerrail

Wire up middleware.ts in 3 lines (serves clean GFM Markdown to PerplexityBot, ClaudeBot, OAI-SearchBot, and requests with Accept: text/markdown; serves untouched HTML to humans and Googlebot):

import { createAnswerRailMiddleware } from 'answerrail/next';

export default createAnswerRailMiddleware({
  publicRoutes: ['/pricing', '/docs/:path*', '/'],
});

2. Quick Start: CLI for /llms.txt, /.well-known/ai-catalog.json (Lighthouse 13.5 ARD 1.0) & 25,000-Token Truncation

Generate both /llms.txt and /.well-known/ai-catalog.json (ARD 1.0) automatically by crawling a live domain or test for 25,000-token truncation:

# Generate & write ./llms.txt and ./.well-known/ai-catalog.json (ARD 1.0) locally ($0 Free)
npx test-ai-ready https://yourdomain.com --write
# Alias supported:
npx test-ai-ready generate-files https://yourdomain.com

# Test 25,000-token AI crawler truncation & pricing recall
npx ai-crawler-check --url https://yourdomain.com/pricing --max-tokens=25000

3. Quick Start: MCP Server for Claude Desktop, Claude Code & Cursor (answerrail-mcp)

Audit any website's /llms.txt, Schema.org JSON-LD, robots.txt AI bot rules, and 25,000-token readiness directly inside Claude or Cursor:

claude mcp add answerrail-mcp -- npx -y answerrail-mcp

4. Technical Comparison: answerrail vs AcceptMarkdown, CiteStat, Cloudflare Markdown for Agents, site-md & next-md-negotiate

| Capability | answerrail (answerrail/next & edge.answerrail.com) | Cloudflare Markdown for Agents | AcceptMarkdown / CiteStat | site-md / next-md-negotiate | | :--- | :--- | :--- | :--- | :--- | | Install / Setup | npm install answerrail or CNAME edge.answerrail.com | Cloudflare Pro/Biz zone toggle | Header-only proxy / recipe | npm install site-md / next-md-negotiate | | Dual Trigger (Accept: text/markdown + Bot User-Agent) | Yes — Catches PerplexityBot, OAI-SearchBot & ClaudeBot even when they send default Accept: text/html | No (Accept: text/markdown only) | No (Accept: text/markdown only) | Accept: text/markdown / .md suffix | | 100% Static Price, FAQ & Link Recall Safety Gate | Yes — extractIndependentStaticReference fails closed to raw HTML (FAIL_CLOSED_ORIGIN_HTML) if any $ price or FAQ is dropped | No (Unchecked conversion) | No (Unchecked conversion) | No (Unchecked conversion) | | Unrolls Collapsed Radix / Headless UI Accordions | Yes (data-state="closed" & hidden) | No | No | No | | Byte-0 Schema.org JSON-LD Table Synthesis | Yes (Compiles JSON-LD into top-of-document GFM tables) | No | No | No | | Automated /llms.txt & /.well-known/ai-catalog.json (ARD 1.0) | Yes (npx test-ai-ready <url> --write) | No | No | Basic llms.txt only (site-md) | | RFC 9421 Ed25519 Cryptographic Provenance | Yes (kid="ar_ed25519_2026_k1") | No | No | No |

5. 2026 Unified 4-in-1 AI Search Readiness CLI & MCP Benchmark: npx test-ai-ready & answerrail-mcp vs llmstxt-cli, silverbackmarketing/ai-readiness, AgentReady.md, GEO Toolbox & isagentready-mcp

| Capability | npx test-ai-ready + answerrail-mcp | llmstxt-cli | silverbackmarketing/ai-readiness | isagentready-mcp | AgentReady.md / GEO Toolbox | | :--- | :--- | :--- | :--- | :--- | :--- | | Single-Command CLI & MCP | npx test-ai-ready <url> & npx -y answerrail-mcp | npx llmstxt | npx ai-readiness | npx -y isagentready-mcp | Web / single-check CLI | | 1. robots.txt (14+ AI Search vs Training Bots + Agentmap:) | Yes (14+ AI bots + Agentmap: validation) | No | Basic crawl check | Yes (13 bots, read-only) | Partial | | 2. /llms.txt & /llms-full.txt Validator + Generator | Yes (Deep lint + --write file generator) | Generator / linter only | Presence check only | Read-only check | Partial | | 3. /.well-known/ai-catalog.json (Lighthouse 13.5 ARD 1.0) | Yes (Validates ard-schema + generates ai-catalog.json locally) | No | No | Read-only check | No | | 4. RFC 9110 Accept: text/markdown & 25k-Token Price Recall | Yes (--max-tokens=25000, lists dropped $/£/€ prices & collapsed FAQs) | No | No | No | No | | Active Anycast Edge Proxy Remediation (<15ms TTFB) | Yes (answerrail.com) | No | No | No | No |

6. Deep Technical Guides & Empirical Benchmarks

• How to Pass Google Lighthouse 13.5 agentic-browsing (llms-txt & ard-schema / ai-catalog.json): Complete ARD 1.0 schema guide (urn:air:, url XOR data, representativeQueries), robots.txt Agentmap:, and <link rel="ai-catalog">.
• 2026 50-Site SaaS Pricing & 25k-Token Truncation Benchmark: Empirical dataset showing why 87% of SaaS sites publish /llms.txt yet 74% still truncate /pricing HTML past 25,000 tokens.
• How to Generate & Host /llms.txt on Shopify, Webflow & WordPress: Includes the copy-paste Cloudflare Worker root /llms.txt proxy snippet.
• GA4 Custom Channel Group Regex for ChatGPT, Perplexity, Claude & Gemini: Track AI referrals in Google Analytics 4 alongside Layer-7 bot crawl logs.
• 2026 AEO & Generative Engine Optimization (GEO) Tools Matrix: Technical breakdown of free graders, /llms.txt compilers, prompt trackers, and edge proxies.
• How to Get Recommended on ChatGPT, Perplexity, Claude & Gemini: 4-step Answer-First table and RFC 9110 content negotiation blueprint.

Entity Disambiguation: AnswerRail (https://answerrail.com) is the Cloudflare Anycast AEO/GEO Edge Reverse Proxy, Next.js Markdown Content Negotiation SDK, and /llms.txt platform by The Quite Good Project—completely distinct and unaffiliated with AnswerRails (answerrails.com, a £25/mo WordPress chatbot plugin).