How to convert a PDF to Markdown
- Drop the PDF. Read in the browser, not uploaded.
- Keep the format on Markdown. Word, HTML, EPUB and CSV are in the same picker.
- Download the .md. Paste it wherever Markdown goes.
Why Markdown rather than plain text
A plain text dump of a PDF loses the one thing that makes a long document navigable: which lines were headings. Markdown keeps that in a form every tool understands, while staying readable on its own. It also keeps lists as lists, and rejoins paragraphs that the PDF had broken into fixed lines, which is what makes the text usable rather than a column of fragments.
If you want only the words with no structure at all, extract the text. For editing, Word. For publishing, HTML.
How the structure is worked out
Knowing how the structure is inferred explains both why the output is good and where it goes wrong. A PDF records characters drawn at given positions in a given font and size, with nothing marking a line as a heading. The structure you see when you look at a page is something your eye reconstructs, and it has to be reconstructed here too.
So the converter measures. It works out which size most of the document's
body text is set in (by volume of characters, not by counting lines, so a
few large headings can't skew it) and treats that as the baseline.
Anything set noticeably larger becomes a heading, with bigger jumps mapping
to higher levels. Short bold lines that don't end like a sentence are read
as run-in headings. Lines beginning with a bullet character, or with
1. or a), become list items. Consecutive lines in
the same size and style are rejoined into one paragraph, which is what
turns a column of fixed-width fragments back into prose you can reflow.
Documents with consistent typography convert well, because the inference has real signal to work with: reports, papers, books, anything made from a template. Where it struggles is where the visual design carried meaning the measurements can't see: a heading distinguished only by colour, a document where body text and headings are the same size, or a title page whose type is large for decorative reasons and arrives as a heading.
What doesn't survive the trip
Markdown is a simpler format than PDF, so some things have nowhere to go:
- Complex tables. A PDF table is lines and text at coordinates, with no notion of rows and cells. Simple grids often come through legibly; merged cells and nested headers do not. For tabular data specifically, PDF to Excel is aimed at the problem.
- Multi-column layouts. Academic papers and newsletters are the classic difficulty; text is extracted in the order it sits on the page, which can read across the columns rather than down them.
- Images. Markdown references image files rather than embedding them, so pictures are not carried over. Pull them out with PDF to images if you need them.
- Footnotes, headers and footers. These are page furniture with no Markdown equivalent, and tend to arrive inline where they fell.
- Scanned pages. No text layer means nothing to convert. OCR the scan first.
Treat the output as a strong first draft rather than a finished document: for most files the tedious 95%, retyping the prose, is done, and what remains is fixing headings and tables in a text editor.
Markdown as model input
A growing reason to convert is feeding documents to a language model or a retrieval system, and Markdown is a better carrier than plain text for it. Headings survive as headings, so a chunker can split on real section boundaries instead of arbitrary character counts, and a model reading the result can tell a section title from a sentence. Lists stay lists. Paragraphs arrive whole rather than as line fragments, which materially improves retrieval quality, since a paragraph broken into eleven lines embeds as eleven fragments of a thought.
It is also compact. Markdown carries structure with a handful of extra characters, where HTML spends tags on it, so more of the document fits in a given context window. And since the conversion happens on your device, a confidential document can be prepared for a model you run locally without passing through anyone else's server on the way.
Common uses for PDF to Markdown
- Notes and wikis: get a document into Obsidian, Notion or a team wiki with its headings.
- Static sites: Markdown is what most site generators want.
- Feed it to a model: structured Markdown works far better than a raw text dump.
Frequently asked questions
What is the Markdown for?
Notes apps, static site generators, wikis, git repositories, and increasingly as input to language models, which handle Markdown far better than they handle a PDF's raw text dump.
What does it convert?
Headings become # levels based on the type size, bullet lists become - items, and paragraphs are rejoined across line breaks with hyphenated words repaired. Everything else becomes plain text.
Are tables converted to Markdown tables?
No. Detecting a table reliably needs the same geometry work the CSV mode does, and a badly guessed Markdown table is worse than plain lines. Use the CSV mode for tables.
Is my document uploaded?
No. The text is read and the Markdown is written in this tab.