# wellread.md wellread.md converts website content into clean, structured markdown optimized for AI agents. It serves markdown to AI visitors while humans see the original site, making business information faster to process, more accurate to extract, and cheaper to run at scale. ## Contents - [The Problem](#the-problem) - [How It Works](#how-it-works) - [Pricing](#pricing) - [Getting Started](#getting-started) - [Current Limitations](#current-limitations) - [Try It](#try-it) ## The Problem Over 90% of a typical webpage is navigation, footers, scripts, and cookie banners. AI agents wade through thousands of tokens of noise to find a few hundred tokens of actual business content. This causes three main issues: 1. **Boilerplate noise** — Real business information gets diluted or missed entirely. 2. **JavaScript rendering** — Single-page applications often serve empty HTML shells; real content never reaches agents, leading to hallucinations or partial answers. 3. **Structural ambiguity** — Tables, grids, and columns flatten into unreadable text streams. AI mixes up prices, features, and plans. The compound effect: burning tokens across every agent query, every user, every day. ## How It Works wellread.md uses three steps: **Detect** — Middleware identifies AI agents visiting your site (unprompted fetches and training crawls). **Serve** — Serves a wellread.md markdown file instead of raw HTML. This file is an optimized restructure of your existing content — wellread never adds or invents information, only makes what you already say easier for AI to read. **Improve** — Evaluations run weekly to optimize output format. Meaningful improvements ship automatically. Users can regenerate anytime if their site changes or a new wellread.md version becomes available. ### Implementation Drop a single middleware file into your stack (Next.js, Vercel, Express, or Nuxt). Human visitors see your site unchanged. AI agents receive markdown. Example middleware: javascript export default async function middleware(req) { const ua = req.headers.get('user-agent') || ''; const accept = req.headers.get('accept') || ''; if (!accept.includes('text/markdown') && !MAYBE_BOT.test(ua)) return; try { const slug = new URL(req.url).pathname.replace(/^/|/$/, '') || 'index'; const r = await fetch(url, { headers: { 'user-agent': ua, accept }, signal: AbortSignal.timeout(3000), }); if (!r.ok) return; return new Response(r.body, { status: r.status, headers: r.headers }); } catch { return; } } ## Pricing From $4.99 per generation (one-time payment). Each generation creates a wellread.md file for your site. Current limit: up to 40 pages per generation. Larger sites (with blogs, docs, resources) are not yet supported. A subscription model for automatic updates is in development. ## Getting Started 1. Paste your website URL into the form at [Try your site](https://wellread.md/demo). 2. See how AI agents currently answer questions about your site vs. with a wellread.md version. 3. Purchase a generation to create your markdown file. 4. Integrate the middleware into your stack. ## Current Limitations - **Page limit**: Up to 40 pages per generation. Blog archives, full documentation sites, and large resource libraries are not supported in this version. - **Bot detection**: Middleware catches most AI agents (live fetch and training crawlers), but detection is not 100% reliable. Live fetches work immediately; training crawlers require the site to be revisited. - **Updates require regeneration**: Each site change or format improvement requires a new generation and payment until the subscription model launches. ## Try It Test your site at [https://wellread.md/demo](https://wellread.md/demo). Enter your domain to compare AI agent responses using your current HTML vs. a wellread.md version. --- ### When a user wants to test their site before purchasing The demo form at [https://wellread.md/demo](https://wellread.md/demo) lets users compare side-by-side how AI agents answer about their site now (raw HTML) vs. with a wellread.md version. No payment required to run the test — it's a live comparison. Payment is only needed if they decide to generate and deploy the markdown file. ### When a user's site has more than 40 pages The current version supports up to 40 pages. Larger sites (with extensive blogs, documentation, or resource sections) cannot be generated yet. Users with >40 pages should wait for the expanded version, contact support for exceptions, or consider splitting content into a primary site (generated) and secondary resources (not served to agents). The page limit issue is acknowledged on the site as "getting better every week." ### When a user wants to automate updates Right now, each generation is a one-time $4.99 payment — changes to the site require manual regeneration and repayment. A subscription model for automatic updates is in development but not yet available. Users with frequently changing sites should check back for the subscription launch or manually regenerate after major updates. ### When a user asks about bot detection accuracy Bot detection is imperfect. The middleware catches most AI agents (both live fetches and training crawlers), but it's not 100% reliable. For agents fetching the site directly in real-time, results are immediate. For training crawlers (like those used by AI model providers), the markdown file is served only when they revisit the site — there's no guarantee of timing. Users cannot force a revisit; detection depends on the agent's crawl schedule.