The recognised text,
next to the page it came from.
Scanned material is read with layout-aware recognition, and the page image is kept beside the text. A citation shows both: the words and the scan they were read from.
LLX-HANDLER-IMG · R3 · 10 formats
Three things a text-only OCR pass loses.
Layout and reading order
Columns, tables and stamps are recognised as structure, so a scanned table is not read as one long sentence.
The page image
The scan is kept and shown with every citation, so a reader can check the recognition against the page.
Recognition confidence
Uncertain words are marked as uncertain rather than silently guessed — an unreadable stamp stays unreadable.
The questions scans usually decide.
What a signed original says
A signature block or handwritten note is cited to its position on the page, with the image attached.
AuthenticityWhether a stamp or date is legible
Low-confidence recognition is reported, so a date nobody can actually read is never quoted as fact.
ReliabilityWhat a paper-only file contains
A scanned bundle becomes searchable without being retyped, and every hit opens on its page.
RetrievalRead in place, in the format it arrived in.
The file is never converted or re-saved. It is read where it lies, and the handler records a SHA-512 of the bytes it read — so a passage cited today can be replayed against the same file years from now.