Converting PDF to Word: what converts well, what doesn't, and how to fix it

Word files describe intent: this is a paragraph, this is a heading, this table has three columns. PDFs describe placement: put this glyph at x=72, y=690. Converting from the second to the first means guessing the intent back, and that is why no converter — including ours — gets every document perfect. It helps to know where the guessing is easy and where it is hard.

What converts cleanly

• Documents born digital from Word, Google Docs, LibreOffice or LaTeX: consistent fonts, real text, simple flow. Paragraphs, headings, lists and basic tables come back editable and close to the original.
• Single-column reports and letters. Reading order is unambiguous.
• Simple tables with ruled borders. Lines tell the converter where the cells are.

What is hard

• Scanned documents. There is no text to convert — only an image. Run OCR PDF first, then convert; the result will be as good as the recognition.
• Multi-column magazine layouts. The converter has to decide whether two blocks of text are two columns or two paragraphs side by side. Expect some reflow.
• Text boxes, sidebars and floating figures. Word positions these with anchors that don't exist in the PDF, so they may land in slightly different places.
• Tables without borders. Alignment alone is a weaker cue than lines, and merged cells make it worse.
• Unusual fonts. If the font is embedded as a subset without a name Word knows, the converter substitutes a similar one and line breaks shift.
• Forms and annotations. Fillable fields become static text; comments and highlights are dropped.

How PDFCraft converts

PDF to Word runs on the server using pdf2docx (built on PyMuPDF), which reconstructs paragraphs, tables and images from the page geometry. If a document defeats that engine, LibreOffice's PDF import is used as a fallback so you always get a file rather than an error. The output is a standard .docx that opens in Word, Google Docs, Pages and LibreOffice.

For the reverse direction, Word to PDF is far more reliable — LibreOffice renders the document exactly as a print would.

Getting a better result

1. OCR scans first. Always. Conversion works on the text layer only.
2. Convert only the pages you need. Use Split PDF or Extract pages first; less content means fewer chances for layout mistakes and a faster job.
3. Expect to tidy headers and footers. They are often converted as ordinary paragraphs at the top and bottom of each page. Deleting them takes seconds.
4. Fix fonts in one go. Select all in Word and set the font family once; it restores the original line breaks more often than not.
5. If you only need the words, use PDF to Text instead. It is instant, keeps the reading order and skips the layout altogether.
6. Keep the PDF. The converted document is a new file; the original remains the reference.

Conversion is a tool for continuing work on a document, not a way to get pixel-perfect fidelity. Used with those expectations, it saves hours of retyping.