A scanned PDF is really a photo album: every page is one image, and there is not a single selectable character inside. That is why "convert to Word" on a scan produces a document containing one giant picture per page. The missing step is OCR - optical character recognition - which reads the picture and reconstructs real text. Here is how the pipeline works and how to get the best accuracy out of it.
What OCR actually does
OCR engines run three passes over each page image:
- Detect - find regions of text and separate them from images, stamps and signatures
- Recognize - classify each character, using a language model to correct likely mistakes ("rn" → "m")
- Rebuild - output the text with font sizes, positions and reading order, so the Word file keeps the original structure
Modern engines handle clean printed text at 95-99% accuracy. Handwriting, faint faxes and heavily skewed pages are where accuracy drops.
How to prepare a scan for best results
- 300 dpi is the sweet spot - below 200 dpi characters start merging; above 400 dpi you gain nothing but bigger files
- Straighten the page - more than about 5 degrees of tilt and lines start to break incorrectly
- Avoid photos of documents taken at an angle - use a scanner or a scan app that applies perspective correction
- Pick the right language - recognition quality drops sharply if the engine expects English and gets German, or vice versa
Converting with WhizPDF
Upload the scanned PDF to PDF to Word - WhizPDF detects scanned pages automatically and routes them through OCR, so you do not need to pick any mode. The result is a .docx with real, editable text. Free accounts include OCR conversions every day; heavy OCR use is what the Pro plan is for.
What to check after conversion
- Names and numbers first - OCR rarely fails in ways you notice; it confuses 0/O, 1/l, 5/S inside plausible-looking words. Proofread anything that will be printed or filed
- Tables - borders drawn on the scan help the engine; borderless tables may come through as plain text
- Signatures and stamps - they are preserved as images, which is what you want
When the source is a photo, not a scan
If all you have is a phone photo, run it through file scanning first: it flattens perspective, boosts contrast and crops to the page edges. A corrected photo converts dramatically better than the raw shot.
The rule of thumb: OCR quality is bounded by scan quality. Two minutes of preparation beats any amount of post-conversion cleanup.