Which formats convert cleanly, which do not, and why. Updated July 2026.
Copy-paste turns a document into a wall of text: the headings become paragraphs, the tables become tab-separated soup, and the lists lose their nesting. A converter keeps that structure — but how much it keeps depends far more on what you are converting from than on which converter you pick.
The dividing line
A Word file knows what a heading is. It stores structure, and a converter reads it. A PDF stores where the ink goes — the structure was thrown away when the file was made, and every converter has to guess it back from the layout. That single distinction predicts most of the frustration people have with this.
We ran the round trip people actually do, with Pandoc 3.9.0.2 on macOS 26.5.2: a table-heavy HTML document to Word, then Word out to GitHub-flavoured Markdown.
Nothing was lost, including through a Word file in the middle. So "tables break in conversion" is not quite right — clean tables in structured documents come through fine. What genuinely breaks is merged cells, nested tables, and any table living inside a PDF, where there was no table structure left to recover.
Structured documents
Free and open source, and the converter everything else is measured against. pandoc report.docx -o report.md. Dozens of formats in both directions with deep control over the output flavour, and a manual worth reading. Its PDF support runs one way: Pandoc writes PDFs, it does not read them.
Free and open source, from Microsoft. A Python CLI built to turn Office files into Markdown for LLM consumption. If the conversion happens inside a script and a model is the reader, this is the purpose-built choice.
€15, once. Drop PDF, Word, PowerPoint or HTML — several at once — and each becomes its own Markdown file. Tables come through as Markdown tables, images inside a document are extracted to a folder beside the file and re-linked, and a pasted URL is converted through reader mode with navigation and ads stripped. Everything runs on your Mac. macOS 14+.
PDFs
If you are converting a handful of ordinary PDFs, any of the tools above will do. If PDFs are the work — research papers, scanned reports, multi-column layouts — the layout-aware tools will beat a general converter, including ours, and we would rather point you at them than take your money for the wrong job.
Both are terminal tools with a setup cost. That cost is worth paying exactly when manual cleanup of the output would take longer, which for a stack of papers it will.
Two checks
A converted document that referenced embedded pictures reads badly with dead links. Some tools extract images to disk and re-link them, some silently drop them. Convert one image-heavy document and look at the output folder before trusting any tool with the rest.
For a single non-sensitive file, a browser converter is free and instant and hard to argue with. For a contract, a report or anything internal, the document travels to someone else's server to be converted. That is the whole trade.
PX7 Digital makes PX7 Markdown Drop Converter, so this page is not neutral. The Pandoc round trip above was run on this machine on 31 July 2026 with the versions named; details for the other tools come from their own repositories and sites on the same date. Four of the tools here are free, and for difficult PDFs they are the better answer. Tell us if something is out of date.