What converts cleanly — and what does not
A PDF exported from Word, Google Docs, or InDesign still has Unicode text in content streams. We read those runs and write them into a Word document with images as pictures. That is a good conversion.
A phone photo of a paper contract is a picture sitting on a page. There is almost no text to extract. We flag it (character count near zero) and can still give you a .docx that holds page images, or run English OCR in this tab. OCR is slower and imperfect on stamps, handwriting, and low-contrast scans. It is honest about that.
Multi-column magazines, text on paths, and tables drawn as lines plus separately placed numbers will need a tidy-up in Word. No browser converter is Adobe Acrobat’s reflow engine. The difference here is the file never uploaded.
Scanned PDFs
If you see the scanned warning, do not expect a perfect contract in Times New Roman. Run OCR if you need to search; otherwise keep the visual Word file and edit in a copy. For a 40-page scan, OCR is happier on a laptop than on a phone.