how to convert a pdf to markdown for ai (3 free ways).

elisabeth hitz · july 14, 2026 · 4 min read

the fastest way to convert a pdf to markdown for ai is to skip the pdf entirely and find the original doc, and if you cannot, run the file through a free converter like microsoft markitdown or hand it to claude once with a single clear prompt and save what it gives you back. all three routes are free. the point is the same: give the model clean text with real headings, lists, and tables instead of a scanned page it has to squint at.

why bother? because a pdf is a layout format, not a text format. it describes where ink sits on a page, not what the words mean. when the model pulls text out of it, columns collide, tables scramble, and headings become plain sentences. markdown fixes that by carrying the structure in plain characters. this is the same reason markdown beats pdf for claude in the first place, and it is why your ai is not working as well as it should when you feed it raw files.

route one: get the source doc

the best pdf is the one you never made. most pdfs started life as a google doc, a notion page, a word file, or a slide deck. if you have access to that original, use it. copy the doc, paste it into a plain text or markdown file, and you are done. no conversion, no cleanup, nothing lost.

use this route when the file is yours or your team's: a proposal, a policy, a brief, meeting notes, an sop. check the drive folder, the notion workspace, the shared inbox. nine times out of ten the editable version is sitting right there and you were about to convert a photocopy of it.

route two: a free converter

when the original is gone (a downloaded report, a vendor spec, a contract someone sent you), run the file through a free converter. two honest options:

• microsoft markitdown is a free, open-source tool built for exactly this. it turns pdfs, word docs, powerpoints, and more into clean markdown. if you are not technical, see the walkthrough in markitdown, explained for people who don't code.
• a one-click web converter works when you just want it done in the browser. paste or upload, copy the markdown out. good enough for a single file you need right now.

use this route for born-digital pdfs, the kind where you can already select the text with your cursor. if the pdf is a scan or a photo, converters will choke, because there is no text layer to read. those need optical character recognition first, which is a different job.

route three: let claude do it once

the laziest route is often the best. hand the pdf straight to claude and ask for markdown. use a prompt like this, word for word:

convert this to clean markdown, keep every heading, list and table, return only the markdown

then here is the part people miss: save the output. paste what claude gives back into a file and keep it. that clean markdown is now your source of truth. next time you need the model to work with that document, you feed the markdown, not the pdf, and you never pay the conversion tax again. do it once, reuse it forever.

use this route when you want the model's judgment on messy layouts, or when the file is small and you would rather not touch a tool at all.

always verify a couple of values

no converter is perfect, and claude can misread a busy table. before you trust the markdown, spot-check it. pick two or three specific things, a total in a table, a date, a line item, and confirm they survived the conversion intact. numbers are where these tools slip most, so that is where you look. a thirty second check saves you from building work on a value that got mangled in transit.

which route wins depends on what you have. the source doc if it exists, a free converter for born-digital pdfs, and claude for the messy one-offs you want to save and reuse. all free. all cleaner than pasting a raw pdf and hoping.

faq

do i need to pay for a tool to convert pdf to markdown?

no. the source-doc route costs nothing, microsoft markitdown is free and open-source, and claude can do the conversion inside a normal chat. there is no paid tool you need to buy for this.

what if my pdf is a scan and the text won't select?

a scan has no text layer, so plain converters produce garbage. it needs optical character recognition first to turn the image into real text. claude can often read a clear scan directly, but verify the output closely because errors are more likely.

why not just paste the pdf straight into claude every time?

you can, but you re-pay the conversion cost on every message and risk the layout scrambling differently each time. converting once and saving the markdown gives you a clean, stable input you control, which is the whole habit worth building.