Free Word (DOCX) to Markdown Converter
Convert Microsoft Word .docx files to clean Markdown for ChatGPT, Claude, Gemini. Preserves headings, lists, bold, italic, links. 100% in-browser via Mammoth.js.
Drop a file here or
.docx (Office 2007+)
Benefits
Headings 1-6, paragraphs, bullet/numbered lists, bold, italic, links, inline images all map to standard Markdown.
DOCX internally is verbose XML. Markdown is 5-10× smaller — feeding a 50-page Word report to Claude saves $0.50+.
Mammoth.js runs entirely in-browser. No server, no logs, no storage.
How to use
- 1Drop a .docx file or click to choose from your device.
- 2Mammoth.js parses the OOXML and converts to Markdown in seconds.
- 3Review the output + token estimates for 4 AI models.
- 4Copy or download .md, or click 'Open in ChatGPT' to auto-paste.
What is DOCX to Markdown?
DOCX (Office Open XML) has been Microsoft Word's format since 2007 — a ZIP containing many XML files describing content, styles, and images. A 1 MB DOCX can hold ~50 KB of plain text, with ~95% being XML overhead. Feeding raw Word content into ChatGPT (drag-drop or copy-paste) usually loses structure and wastes tokens on XML.
This tool uses Mammoth.js — a 5k-star DOCX → HTML/Markdown library that works entirely offline. Mammoth analyses the OOXML, recognises styles (Heading 1-6, Quote, Code), and maps them to HTML/MD equivalents — preserving the original Word structure for AI to read.
Use cases: students feeding theses into Claude for Q&A; office workers converting contracts/reports into AI-friendly format; devs migrating legacy documents into Markdown knowledge bases (Notion, Obsidian, GitBook).
- ✓Word style Heading 1-6 → # ## ### #### ##### ###### Markdown
- ✓Bullet and numbered lists (multi-level) → - and 1.
- ✓Bold, italic, underline → **bold** *italic* (underline becomes italic since Markdown lacks it)
- ✓Hyperlinks → [text](url)
- ✓Simple tables → GFM tables
- ✓Inline images →  — toggle to strip for compactness
When to use it
Convert monthly/quarterly reports to Markdown then paste into Claude for Q&A — save 80% of tokens.
Students with 50-100 page thesis convert to MD then ask Claude for chapter summaries.
Notion's paste-as-markdown picks up structure better than direct .docx import (which often loses formatting).
Convert SOPs/policies from Word to Markdown for Confluence, GitBook, or RAG.
Convert lesson docs from Word to MD then feed Claude to generate exercises or alternate explanations.
How it works
Mammoth.js reads OOXML node-by-node, using style mapping to recognise 'Heading 1' (Word style) → '# ' (Markdown), 'Heading 2' → '## ', etc. The default mapping covers standard Word styles; documents with custom styles may need custom maps (not yet exposed in UI).
Tables are converted to HTML <table> → then through Turndown to GFM tables. Note: complex tables with merged cells (rowspan/colspan) are degraded to flat tables in Markdown — this is a Markdown format limitation, not a bug.
Images in DOCX are extracted as base64 data URLs embedded inline in the Markdown. With image-heavy documents, Markdown size can exceed the original DOCX — toggle 'Strip images' if you only need text for AI.
DOCX → Markdown FAQ
Does .doc (legacy Word) work?
Not in the free tier. Legacy .doc (Word 97-2003) needs LibreOffice to parse — coming in Phase 2 (paid). Currently supports only .docx (Office 2007+).
What about comments, track changes, footnotes?
Comments and track changes are dropped (only final text is kept). Footnotes are preserved at the end of the document as a list.
Are images preserved?
Yes — converted to inline base64 data URLs. Can significantly inflate Markdown size for image-heavy documents. Toggle 'Strip images' if needed.
What about Microsoft Equation math?
Currently kept as text representation, not LaTeX. To preserve proper $...$ math, you'd need an engine like Pandoc — coming in Phase 2.
Can I edit Markdown before copying?
Yes. Output appears in a textarea — edit directly before copying or downloading.