Convert from PDF · guide
What PDF to Markdown does
PDF to Markdown turns the text of a PDF into a Markdown file, rebuilding paragraphs, headings and lists from how the text sits on the page. It is deliberately cautious: where it is unsure whether a line is a heading, it leaves it as plain text rather than guess. Useful for notes, documentation, and giving a document to an AI assistant as text you have read first. The PDF is never uploaded.
How to convert PDF to Markdown in your browser
- Choose a PDF with selectable text Select a PDF whose text you can select. A scanned PDF has no text layer, so run it through OCR PDF first.
- Convert it Select Convert to Markdown. Lines, paragraphs, headings and lists are worked out from where the text sits on each page, and each page is separated from the next by a horizontal rule.
- Check and save Save the .md file and skim its headings and lists, which are inferred from the layout rather than stored in the PDF.
Worked example
Reuse a text-based meeting memo as editable Markdown notes.
- Starting point
- Choose a PDF where a viewer can select the memo text.
- What to do
- Convert, then review the generated headings, lists and reading order in a text editor.
- Result
- Expect a Markdown file containing extracted text structure, not a copy of the printed page layout.
If the result is not right
If the tool says the PDF has no text layer, the memo is most likely made of scanned images. If only some pages are missing from the result, they may be scanned too. Run OCR first and proofread the recognised text before converting again.
Limits worth knowing
- Multiple columns, unusual spacing and decorative headings can lead to an incorrect reading order.
Questions and answers
Can I use PDF to Markdown to give a document to ChatGPT or Claude?
Yes, and it is a careful way to do it. Converting in your browser gives you plain text with headings and lists intact, which you can read, trim and strip of names before pasting it in, instead of uploading the whole file with everything it contains. Markdown is also compact: the words and their structure, without the fonts, images and layout data that make a PDF large.
What happens with a scanned PDF?
A scanned page is a picture with no text in it, so there is nothing to convert. If every page is a scan, the tool says so plainly instead of returning an empty file; run the scan through OCR PDF first, then convert the result. In a document that mixes typed and scanned pages, the scanned pages are skipped without a note and the file holds only the typed pages, so check that every page you need is there. If a scan arrives with a poor text layer already baked in by the scanner, the notes warn that it reads like bad OCR, because those misreadings are what the Markdown will contain.
Why are two-column pages and tables jumbled?
Because lines are read straight across the page. Text at the same height is joined into one line, so on a two-column page the left and right columns interleave line by line, and a table comes out as rows of run-together cells. For a two-column paper, PDF to Word keeps each line where it was drawn, so the columns stay side by side; for tables, PDF to Excel puts each value in its own cell.
Does it keep bold, links and images?
No. The output is plain Markdown: headings marked with #, bullet and numbered list items, and paragraphs. Bold is not detected and is not written out as bold; a link keeps its words but loses its address; images are left out entirely. To get the pictures as well, run the same PDF through Extract images from PDF.
Why are page numbers and running headers mixed into the text?
Each page is converted on its own and the pages are joined with a --- rule, so whatever is printed on a page, including its header, footer and page number, stays where it falls. A word hyphenated across a line break also stays split, as in inter- national. Find and replace in any text editor clears both quickly.
How does it decide what is a heading?
By size, compared with the body text on the same page. A short line at least 1.6 times the body size becomes #, 1.32 times ##, and 1.15 times ###. A line of more than 14 words is never a heading. Bold is not read, so a sub-heading set at or near body size stays plain text and can run into the paragraph below it.