Printed and typewritten text
Reads best. A smudge, a stamp or a handwritten note can leave a word uncertain, and a poor scan can be read wrongly without anyone noticing.
Guide
A scanned lease, deed or contract is a picture of a page. It cannot be searched or edited until someone turns the picture back into words and structure. Firms have paid people to retype them, or used basic converters that left the page hard to edit.
This page sets out what a scan actually contains, the three ways firms deal with one today, what LexForge returns, and what you should still check by hand.
Last updated
What is in a scan
Optical character recognition (OCR) can read printed and typewritten text from a scan. It cannot see that a line is a heading, that a paragraph is clause 1.1, or that a note at the foot of the page belongs with the body text. Those things have to be rebuilt before the document is useful in Word again.
Reads best. A smudge, a stamp or a handwritten note can leave a word uncertain, and a poor scan can be read wrongly without anyone noticing.
Not in the file. A converter that keeps the page as positioned boxes leaves them for you to redo by hand.
Structure that has to come back as editable Word structure, not as a picture of the page.
Three ways firms do it today
Careful retyping is accurate but slow, and every retyped word is a new chance for a mistake. Nothing marks the words a tired reader misread.
Fast, but pages may remain as positioned boxes, damaged text is left in the page, and whether the wording is checked varies by tool. See the comparison.
OCR reads the words, AI rebuilds headings, numbering and structure, the draft is checked against the source reading, and uncertain words are flagged for a person to review rather than guessed.
What LexForge returns
Optional. A Word template or a well-formatted document teaches LexForge your headings, numbering and footnotes. How House Styles work, and what a house style is.
A scanned PDF or a Word (.docx) file, up to 100 MB; a PDF can contain up to 500 pages. Documents are stored and read in the UK, structure analysis runs inside the EEA, and nothing you upload is used to train AI models. Security page.
Headings, numbering, definitions and tables come back as editable Word structure. Footnote text is kept and returned as editable text in the body rather than as Word's linked footnotes. Every draft is checked against the words read from the source.
Before you rely on it
Words the reading leaves uncertain are highlighted in the draft for a person to check rather than guessed. The wording check confirms that the draft matches what was read from your file; it cannot prove that a poor scan was read correctly, so a smudged word can be consistently wrong in both.
AI can make mistakes. LexForge uses AI for layout and reads words with OCR. Review the finished document before relying on it.
Source scans leave active storage two hours after upload, and finished files stay for the retention period your firm sets, from seven to 30 days.
See it on a real page
Fictional specimen. Every name, party and clause is invented, and no client data is used.
Try it on your own scan
Approved pilot users start with 300 one-time pages. No card required.
Start your free pilot →