The text layer of digitally generated PDFs is generally clean and extractable, while real scanned documents (handwritten annotations, low-quality tilted and stained scans, checkbox and signature areas, cross-page broken tables) are the primary failure scenarios for OCR.