HTML Input — paste, type, or drop a .html file
Markdown Output
HTML is tag-heavy by design. Every <div class="...">, inline style, and attribute costs tokens an LLM wastes on structure rather than content. Markdown stores only the content with minimal syntax — cutting token counts by 40–70% on typical web pages.
Web pages contain navigation, ads, footers, cookie banners, and script noise. This converter strips all of it automatically — nav, header, footer, aside, scripts, and styles — leaving only the semantic content. Feed the clean Markdown directly into your vector store.
Moving content from a CMS, blog platform, or static site to a Markdown-based knowledge base? Convert each HTML page to .md and import it directly into Obsidian, Notion, MkDocs, Hugo, or any Markdown-first platform without manual reformatting.
Conversion happens live in your browser as you type or paste — no waiting, no upload, no API key. Works completely offline after the page loads. Paste HTML from DevTools, a saved file, or a scraper output and get Markdown instantly.
Headings, bold, italic, lists, tables, links, images, code blocks, and blockquotes are all mapped to their Markdown equivalents. The document hierarchy is preserved for accurate heading-based chunking in RAG systems.
Markdown renders natively on GitHub, GitLab, Notion, Obsidian, Confluence, and every AI developer tool. HTML requires a browser or renderer. Markdown is version-control friendly, diffable, and will be readable in any text editor indefinitely.
The converter parses your HTML using a series of ordered regex transformations — no external library required. It first strips noise elements: <script>, <style>, <nav>, <footer>, <aside>, and <form> elements entirely.
Then it maps semantic HTML to Markdown: headings → # syntax, <strong> → **bold**, <pre><code> → fenced code blocks, lists → dashes and numbers, tables → pipe syntax, blockquotes → > lines. HTML entities are decoded to Unicode. Finally, remaining tags are stripped and excessive blank lines collapsed.
All processing runs in memory in your browser tab. Nothing is ever sent to a server.
**bold**, em/i → *italic*[text](url)| Feature | Markdown (.md) | HTML (.html) |
|---|---|---|
| LLM token cost | ✅ Low — minimal syntax overhead | ❌ High — tags, attributes, classes |
| RAG chunking quality | ✅ Clean heading/paragraph structure | ⚠️ Requires HTML parsing and noise removal |
| File size | ✅ Typically 40–70% smaller | ❌ Bloated with tag and attribute overhead |
| Git / version control | ✅ Clean, readable diffs | ⚠️ Diffable but noisy with attributes |
| Editable without tools | ✅ Any text editor | ⚠️ Requires HTML knowledge |
| Obsidian / Notion import | ✅ Native format | ⚠️ Requires conversion or plugin |
| Browser rendering | ⚠️ Requires Markdown renderer | ✅ Native browser support |
Why is Markdown more token-efficient than HTML for LLMs?
Every HTML tag, attribute, class name, and inline style is a token. A single <div class="container mx-auto px-4 py-8"> consumes roughly 12 tokens before any content appears. A full web page with navigation, footer, sidebars, and ad slots can have thousands of structural tokens surrounding a few hundred words of actual content. Markdown's syntax adds almost no overhead — a heading is # text, bold is **text**. The content-to-token ratio is dramatically better.
How do I convert a webpage to Markdown?
In your browser, open the page and press Ctrl+U (or Cmd+U on Mac) to view the page source. Select all, copy, and paste it here. Alternatively, right-click the main content area in DevTools (F12 → Elements), right-click the content element → Copy → Copy outerHTML — this gives you just the article or main content without the surrounding page chrome.
Is my HTML uploaded to a server?
No. All conversion runs entirely in your browser. Your HTML code never leaves your device.
What HTML elements are stripped?
Stripped entirely: <script>, <style>, <noscript>, <svg>, <nav>, <header>, <footer>, <aside>, <form>. All remaining HTML tags that don't map to a Markdown equivalent are also removed, leaving only text content.
Can I use the output directly in a RAG pipeline?
Yes. Download the .md file and pass it to MarkdownHeaderTextSplitter (LangChain) or MarkdownNodeParser (LlamaIndex). These chunk on heading boundaries — so your H2 sections become individual retrievable chunks aligned with the document's actual semantic structure.
Can I import the Markdown into Obsidian or Notion?
Yes. For Obsidian, copy the .md file into your vault. For Notion, use Import → Markdown & CSV. Both apps render headings, bold, italic, links, tables, and lists natively.
How does HTML to Markdown compare to PDF to Markdown accuracy?
HTML to Markdown is generally more accurate. HTML has explicit semantic tags (<h1>, <strong>, <ul>) that map directly to Markdown. PDF requires inferring structure from text position and font size — headings are detected by size ratios, and tables rarely reconstruct well. If your source exists as both HTML and PDF, prefer the HTML conversion.
Does it work with HTML emails?
Yes. HTML emails are often heavily table-based for layout compatibility, but the converter extracts text content, headings, and links. The visual multi-column layout won't reconstruct, but the readable content will be extracted cleanly as linear Markdown text.