Drag-select a sentence in any PDF reader. If nothing highlights, it is a scan.
Pick the editable mode so the server runs text recognition before rebuilding the layout.
Start with the key pages and check them before processing the rest.
PDF is a page description format: it records which glyph goes at which coordinate, plus the lines and rectangles around it. There are no paragraphs, no table structure, no line-spacing semantics. Converting to Word means guessing that structure back, and when the guess is wrong you get jumping font sizes, collapsed tables and dozens of stray text boxes.
A whole page turned into one image means the source was a scan with no text layer. Selectable text that ends up scattered means unusual fonts or a complex multi-column layout. A few characters showing as boxes means the original did not embed its fonts and the converter has no substitute.
First determine whether the PDF is a scan. Then retry with a different conversion mode (layout-preserving vs fully editable). Finally convert only the pages you actually need — whole-document conversion has the highest error rate.
Formulas are usually kept as images and complex ones may need retyping; charts are kept as images rather than redrawn.
No hard limit, but longer documents take longer — process them in chapters.
No. The value is skipping manual retyping; details still need a human check.