OCR
Scanned pages stay in the library as original images. This web session does not invent Arabic from pixels. Selectable PDF text is extracted and translated; image-only pages stay honest and marked as unavailable.