Converts a Word document to Markdown, and it is the easy direction. Unlike a PDF, a .docx states its own structure — the headings are tagged as headings and the tables are tables — so nothing has to be guessed from the layout. Open the converter. Nothing is uploaded.
A .docx is a zip of XML, and that XML says what everything is. A heading carries a style name, a list is a
numbered list with a level, a table is a grid of cells, a hyperlink knows where it points. All of that maps
onto Markdown directly.
A PDF records none of it — a heading there is just some text that happens to be bigger — which is
why PDF to Markdown has to infer structure and occasionally gets it wrong,
and this does not. If you have a choice of source file, use the Word one.
Element
Carried across?
Heading 1–6
Yes — to # through ######
Bullet and numbered lists, nested
Yes, with indentation
Bold, italic, strikethrough
Yes
Tables
Yes, as GitHub-flavoured tables
Hyperlinks
Yes, with the target
Block quotes and code
Yes
Images
No — see below
Comments and tracked changes
No — dropped, not accepted or rejected
Headers, footers, page numbers
No — Markdown has no pages
Fonts, colours, spacing
No — Markdown has no styling, by design
Tracked changes are dropped, and that is worth knowing before you send it
If the document has unaccepted revisions, the Markdown contains neither the insertions nor the deletions
marked as such — it contains the text as the document currently reads. Nothing is lost that was visible,
but the history is gone, and so are comments. For a document under review, accept or reject the
changes in Word first so you know exactly which version you are converting.
Images are not embedded, and there is nowhere for them to go
Markdown references images by path; it cannot hold one. A single .md download has no folder to
put the files in, so pictures are dropped rather than written as links to files that do not exist. If the
images are the point, convert to DOCX or PDF instead, or pull the text out
here and attach the pictures separately.
What Markdown costs in tokens
What the page holds
Plain text
Markdown
Difference
Prose only
53 tokens
52 tokens
1.9% cheaper
Headings and lists
57 tokens
60 tokens
+5.3%
A table
35 tokens
59 tokens
+68.6%
All three together
146 tokens
172 tokens
+17.8%
Those are token counts from our own converter, on a PDF we built for the purpose, measured with the same tokeniser the GPT models use. The pattern is the useful part, and it is not the one usually quoted:
On prose, Markdown is free. It came out 1.9% cheaper than the plain-text dump of the same page, because joining wrapped lines back into paragraphs saves about as much as the heading marks cost.
On headings and lists, near enough free — +5.3% for structure a model can navigate.
On tables it is expensive — +68.6%, all of it pipes and the separator row. And it is the one case where you have no choice, because plain text loses which number sits under which heading.
Those figures come from our PDF converter because that is where we could build a document with a known
shape and measure both outputs of the same file. The Markdown a Word document produces is the same Markdown,
so the trade is the same.
It runs in your browser
The .docx is read and converted in the page, so it is never uploaded. Word documents are where contracts,
reports and HR material live, which makes this the conversion people are most often told not to do on a
random website. Here there is no server copy to worry about, because there is no server in the path.
Frequently Asked Questions
Open the converter, choose DOCX as the input and Markdown as the output, and drop the file in. It converts in your browser and downloads as a .md file. Headings, lists, tables, links and emphasis all come across, because the .docx already records what each of those is.
Yes, when you have the choice. A .docx states its own structure — a heading is tagged as a heading — so nothing is guessed. A PDF only has text at coordinates, so headings and tables have to be inferred from font sizes and spacing, which is usually right but not always. If the Word original exists, convert that.
Images, comments, tracked-change history, headers and footers, page numbers, and all styling — fonts, colours and spacing. Markdown has no way to express any of those. What survives is the structure and the words, which is what you want when the next reader is a language model.
They are dropped, not accepted or rejected. The Markdown contains the text as the document currently reads, with no record of the revisions and no comments. For anything under review, resolve the changes in Word first so you know which version you converted.
Yes — the reverse conversion exists and maps headings, lists, tables and emphasis back into real Word structure rather than flat paragraphs.
No. The file is read and converted by JavaScript running in your own tab, so it never leaves your machine. Word documents tend to hold contracts, reports and HR material, which is exactly the content people are warned not to paste into a random web tool.