The retyping problem nobody enjoys
You've got a PDF — a report, a spec, an old README exported years ago — and you need its content in Markdown for a docs site, a note-taking app, or a README file. Copy-pasting from a PDF viewer usually produces a wall of text with the formatting stripped and line breaks in strange places, leaving you to manually add back headings and structure by hand. For anything longer than a page, that's genuinely tedious work for something that should be automatic.
A PDF-to-Markdown converter exists to skip that retyping entirely — it reads the text and rebuilds a reasonable structural approximation automatically.
PDF has no headings — so how does this guess where they are?
This is the part worth understanding, because it explains both what works well and what doesn't: a PDF file has no built-in concept of "this is a heading" or "this is a paragraph" — it only knows where each piece of text sits on the page and what font size it's drawn at. A converter has to infer structure from that alone, and the reliable signal available is font size: text noticeably larger than the page's median size gets treated as a heading, everything else becomes a plain paragraph.
This is exactly how a human skims an unfamiliar document too — bigger text reads as "more important," which is why the heuristic actually works reasonably well on documents that were designed with a normal visual hierarchy.
Where this genuinely struggles
A document where every line uses roughly the same font size — no visually distinct titles or section headers — gives the converter nothing to work with, so you'll get a Markdown file that's technically correct but flat: every line becomes a plain paragraph with no heading structure at all. Similarly, a scanned PDF (an image of text rather than actual embedded text data) produces nothing, since there's no text layer to extract in the first place — that needs OCR first, not this tool.
Bold or italic emphasis within a paragraph also doesn't carry over — only the heading-versus-paragraph structure is inferred, so a document relying heavily on inline emphasis rather than heading hierarchy will lose that nuance in the conversion.