How to Convert Word to Markdown: 3 Reliable Methods

To convert Word to Markdown, upload the .docx to a converter, run Pandoc, or save it as HTML and convert that. All three keep headings, lists, links, and bold, as long as the Word file uses real Heading styles. Images, tables, and footnotes are where they differ, so this guide shows what each route does with those, based on how each tool documents its behaviour.

Which Word to Markdown Method Should You Use?

Pick by what's in the document and where you're sitting. A browser word to markdown converter is fastest for a single file. Pandoc wins for batches, for getting the images out as files, and for control over the output flavour. The HTML route is the one that handles a legacy .doc or a single section without converting the whole file.

Situation	Best method
One .docx, no install allowed	Method 1: upload to the DOCX to Markdown tool
Many files, or the images matter	Method 2: Pandoc
Legacy .doc, or only one section needed	Method 3: save as HTML, then HTML to Markdown
Output is for an AI pipeline, not humans	Microsoft MarkItDown (noted under Method 2)

Whichever you choose, the reverse direction is a different problem. Our guide to converting Markdown to Word covers going back.

Method 1: Convert Word to Markdown in the Browser

The DOCX to Markdown tool takes a .docx upload and loads the result into the editor with a live preview. It maps Word's Heading 1 to 6 styles to # headings and keeps bold, italic, nested lists, tables with header rows, hyperlinks, and inline code.

For a report with images, a table, a footnote, and a bulleted list, here is what the docx to markdown converter does, per its own tool page:

• Headings, bullets, bold, and links come through as clean Markdown.
• Embedded images are not extracted at the moment. Add the image references by hand after conversion, or use Pandoc when the images matter.
• Tracked changes, comments, custom styles, and macros are stripped, so accept or reject changes in Word first if you want them reflected.
• Footnotes are not covered by the tool's feature list, so check them by hand and re-add [^1] markers where needed.

That's still the quickest route for a one-off file, and the preview shows problems immediately. The limitation is real, though: if the document is image-heavy, Pandoc's --extract-media does that job in one pass.

Method 2: Convert DOCX to Markdown with Pandoc

Pandoc is the standard command-line converter, and its docx reader is mature. Install it, then run one command:

pandoc report.docx -f docx -t gfm --extract-media=./media --wrap=none -o report.md

Each flag matters. -t gfm writes GitHub Flavored Markdown with pipe tables; the default -t markdown writes Pandoc's own flavour, which can use multiline or grid tables and appends {width=...} attributes to images that carry size information. Per the Pandoc manual, --extract-media pulls the images out of the .docx container into the folder you name and rewrites the references to point at them. --wrap=none stops Pandoc re-wrapping paragraphs at 72 columns, which keeps Git diffs sane.

The gfm writer keeps footnotes as [^1] and writes pipe tables, both of which GitHub renders. -t markdown_strict is the wrong choice for this job: the original Markdown.pl syntax has neither tables nor footnotes, so Pandoc has to fall back to raw HTML or plain text for them. Stick with gfm or markdown. Tracked changes are accepted by default; --track-changes=reject discards them and all keeps both versions in spans.

To convert word document to markdown files in bulk, loop in the shell:

for f in *.docx; do
  pandoc "$f" -f docx -t gfm --extract-media="./media/${f%.docx}" --wrap=none -o "${f%.docx}.md"
done

If you'd rather stay in Python, Microsoft's MarkItDown converts .docx with pip install 'markitdown[docx]' (Python 3.10 or newer) and markitdown report.docx -o report.md. Its README says the output is tuned for language models rather than human-readable fidelity, so expect plainer results than Pandoc.

Method 3: Save as HTML, Then Convert to Markdown

Word itself can't save Markdown, but it can save HTML, and HTML converts cleanly. This route also solves two problems the others can't: a legacy .doc file, which Pandoc's input list doesn't include, and a single chapter you don't want to extract by hand.

1. Open the document in Word. To convert doc to markdown, this step is where the old binary format gets read.
2. Choose File > Save As and set the format to Web Page (.htm). Microsoft's file format list describes it as the HTML export; on Windows, Web Page, Filtered produces leaner markup.
3. Open the .htm in a text editor, copy the part you need, and paste it into the HTML to Markdown tool.

Word's HTML is verbose, with inline styles on nearly every element, so the converter strips a lot. Headings still map, because Word writes them as <h1> to <h6>. Our HTML to Markdown guide covers what survives that conversion. For a .doc you plan to convert repeatedly, save a copy as .docx once instead and use Method 1 or 2.

Why Don't My Headings Convert?

This one fact fixes most bad conversions: every Word Markdown converter reads paragraph styles, not what the text looks like. A line set to 16pt bold by hand is still style "Normal", so it becomes a plain paragraph. A line with the Heading 2 style becomes ##. The Pandoc manual states the docx reader "only reads those styles that it can convert into pandoc elements", and browser converters work the same way.

Before converting a document that someone else formatted, spend two minutes in Word:

• Click each heading and apply Heading 1, Heading 2, or Heading 3 from the Styles gallery.
• Turn manually typed "1." and "-" list prefixes into real Word lists with the toolbar buttons.
• Accept or reject tracked changes, and delete comments you don't want in the output.
• If the file is .doc, save it as .docx first.

Two more traps. Word's smart quotes and non-breaking spaces survive every method as Unicode characters; they're harmless in prose but break code samples, so search for them in any command lines. And a dollar sign in the text can come out as \$ from Pandoc's gfm writer, which is correct Markdown but looks odd until you know why.

What Each Method Does with the Hard Parts

Here's how the same kind of document fares through all three routes, based on what each tool's documentation says it does.

Element	Browser tool	Pandoc -t gfm	Save as HTML
Heading styles	# headings	# headings	# headings
Images	Not extracted; add references by hand	Files via --extract-media	Separate files in a folder
Simple table	Pipe table with header row	Pipe table	Pipe table
Merged cells	Span lost, columns misalign	Pipe table can't express spans	Same limit
Footnotes	Check by hand	[^1] footnotes	Plain numbered text
Tracked changes	Stripped	Accepted by default, flag to change	As displayed when saved
Text boxes	Text-based content only	Text extracted, position lost	Often lost

Merged cells deserve a sentence. Markdown pipe tables have no colspan, so a two-column header over three body columns misaligns in every method. Rebuild those tables by hand with our Markdown table guide, or keep them as raw HTML. For extracted image paths that don't resolve, the Markdown image guide explains relative paths; for footnote syntax by platform, see Markdown footnotes.

Clean Up the Converted Markdown in the Editor

Every method leaves a little work. The editor is the place to do it, because the preview shows each fix as you make it. The demo below is loaded with a typical post-conversion cleanup list; on the tool page itself, the upload button sits above the editor. Footnote syntax is one thing this preview does not render, so check footnotes on GitHub or in Pandoc's output instead.

We prefer to run Pandoc first and paste its output into the editor for the cleanup pass, because Pandoc's footnotes and image files need the least fixing. For a document with no images, the browser tool alone is enough.

Word to Markdown FAQ

Converting Word to Markdown comes down to three choices. Use the browser tool for a quick single file, Pandoc for batches and clean image handling, and the HTML export for .doc files or partial documents. Fix the Heading styles first and most of the cleanup disappears. Whichever route you take, finish in the editor so you can see the rendered result while you tidy the tables.