Optical character recognition is pattern matching over pixels. It has no understanding of what a document means, so its accuracy is dominated by things you control before recognition starts: resolution, contrast, straightness and layout complexity.
These guides explain what OCR does to a file, why the same engine can give a perfect result on one page and nonsense on the next, and how to prepare scans so recognition has a fair chance.
OCR reads pixels and writes characters into an invisible layer behind the page image — the picture you see never changes.
Recognition accuracy is set by resolution, contrast and straightness — fix the capture and the results improve far more than changing tools ever will.