Why PDF to Word Conversion Is Fundamentally Lossy

A PDF stores text as positioned characters on a canvas. A Word document stores text as structured paragraphs with styles. Converting between them requires inferring structure from position which is inherently imperfect. The fundamental mismatch PDF text is positioned characters: "Hello" at position (72, 720) "World" at position (72, 700) Word text is structured content: <w:p> <w:r><w:t> Hello </w:t></w:r> </w:p> <w:p> <w:r><w:t> World </w:t></w:r> </w:p> The converter must infer that "Hello" and "World" are separate paragraphs based on their vertical positions. But what if they are two columns? Or a heading and body text? Or a table cell and adjacent content? The positional information alone does not answer these questions. What gets lost Paragraph structure. The converter guesses paragraph boundaries based on vertical spacing and indentation. It is wrong roughly 5-10% of the time, especially with complex layouts. Tables. PDF tables are not tables. They are lines and text at specific p