How to Convert a PDF to Markdown or Plain Text (for Notes and AI Tools)

Plain text loses structure — headings, lists, and emphasis all flatten into the same block of text. Markdown keeps that structure using lightweight syntax that renders correctly in GitHub, Notion, documentation platforms, and static site generators, which makes it a better target than Word for technical or web-based content.

It also beats copying text straight out of a PDF viewer, which tends to bring along odd line breaks, uneven spacing and no heading structure. A converted .md file has consistent formatting that note apps and editors parse cleanly.

The tool extracts the text and reconstructs the structure as Markdown syntax.

Headings and paragraphs convert best. Lists, tables, multi-column layouts and heavily designed pages come through as plain paragraphs and need some manual cleanup, because the converter only tells headings from body text.

The converter reads the text in the PDF and treats larger text as headings and everything else as paragraphs. That works well for reports, articles and notes with clear heading sizes. Documents where everything is the same size come out as plain paragraphs, which you can tidy up in any Markdown editor.

If the converter reports that very little text was found, the PDF is probably a scanned image with no real text in it. OCR PDF can recognise the words in the scan (English text) and give them to you as a .txt file, which you can paste into a Markdown editor and add headings to.

You do not need an Obsidian plugin to bring a PDF into your notes: the converter gives you an ordinary .md file.

The .md file opens in note apps such as Obsidian and Notion, in code editors, and on platforms like GitHub. It is also a compact way to give an AI assistant the text of a document without the formatting overhead of the original PDF.

Some AI chat tools don't support file uploads at all, especially on free tiers. Others do, but process the document more slowly or less reliably than plain pasted text. And if you only need the AI to look at a page or two out of a much longer document, extracting just the text you need is quicker than uploading the whole file and asking it to find the relevant section.

The process is the same regardless of which AI tool you're pasting into afterward.

If you only need the AI to look at a specific section of a long PDF — one chapter, one clause, a few pages — extracting just those pages into a smaller file first keeps the amount of text you're pasting manageable, and keeps the AI's attention on the part that actually matters.

Summarizing a long report before a meeting, asking an AI tool to explain a dense clause in a contract, pulling key numbers out of a research paper, or getting a second read on an essay before submitting it are all common reasons people extract text this way. In each case, the AI only needs the words — not the PDF's layout, page numbers, or formatting — so plain text is the more efficient input.

Extracting the text is entirely local to your browser and doesn't involve any server. What happens after you paste that text into an AI chat tool is a separate matter, governed by that tool's own privacy policy — worth a quick check if the document contains anything sensitive.