Serving HTTP Accept: text/markdown, Content-Signal & /llms-full.txt on Cloudflare Pages (2026 GEO Guide) A developer built GeoReady, a free real-time edge inspector, and an open-source zero-dependency MCP server (yixuanrobot/geoready-mcp) that audits any URL in about five seconds for AI-agent readiness. The tool checks robots.txt rules for bots like OAI-SearchBot, PerplexityBot and ClaudeBot, llms.txt and llms-full.txt presence, JSON-LD structured data, and Content-Signal HTTP headers. The project promotes serving Markdown via HTTP Accept: text/markdown content negotiation, which the developer says cuts payloads from roughly 16,000 to 2,000 tokens. Modern AI coding and shopping agents ClaudeBot , OAI-SearchBot , PerplexityBot , Claude Code waste up to 85% of their context window parsing inline CSS, SVGs, and layout boilerplate when fetching raw HTML. In February 2026, HTTP content negotiation via Accept: text/markdown and the Content-Signal header emerged as the cleanest white-hat Generative Engine Optimization GEO pattern for SaaS and Shopify sites: Accept: text/html receive your normal interactive page; Accept: text/markdown receive 1:1 visible content converted into clean YAML-frontmatter Markdown tables Vary: Accept , cutting payload size from ~16,000 tokens to ~2,000 tokens with zero SEO cloaking risk. We built a free real-time edge inspector — GeoReady https://geoready.aitruelabs.com — plus an open-source zero-dependency MCP server yixuanrobot/geoready-mcp https://github.com/yixuanrobot/geoready-mcp that audits any URL in 5 seconds across: robots.txt OAI-SearchBot , PerplexityBot , ClaudeBot , Google-Extended + firewall HTTP 403 challenge detection /llms.txt & /llms-full.txt index.html fallbacks Organization , Product / SoftwareApplication , FAQPage Content-Signal HTTP Headers ai-train=yes, search=yes, ai-input=yes headers Snippet Drop this into your static output folder's headers file: / Content-Signal: ai-train=yes, search=yes, ai-input=yes Vary: Accept Link: