

When a contract arrives as a scan, the first question is whether there is any readable text inside the file. Some scanners run text recognition on the way out. Some do not, and a file that went through neither is a stack of page images with nothing in it to search. You cannot tell which one you have by looking at it.
Reviewers usually find out in the middle of the job, when a search for the indemnity language across a sixty-page supplier agreement comes back empty. By then someone has spent an hour on a file that was never ready to be reviewed.
Open it and try to select a paragraph. Search for a word you can see on the page. If nothing highlights and the search finds nothing, treat the file as image-only. A locked file or a broken font can fail the same test, and the fix is the same: get a version with real text in it.
Running text recognition puts a text layer over the scan. Adobe's walkthrough of the scan and OCR steps in Acrobat is the short version: enhance the image, recognize the text, and the file comes out searchable, assuming the scan was clear enough to read. Your PDF software probably already does it. If the file is a fax of a fax, ask for a clean copy instead.
This is intake work, and it sits right next to the routing call about where a file goes. A file nobody can search cannot be routed on anything except the requester's description of it.
The mixed file is the one that gets through. The agreement was drafted in Word and exported cleanly. Then somebody printed the pricing exhibit, got it signed, scanned it and stapled it onto the back. Pages 1 through 40 are text. Pages 41 through 48 are photographs.
Anyone who checks the first page calls the file fine, and the exhibit carrying the rates goes through the whole review unread by anything. So the check has to cover the whole file, which is why it should be a step in the process that somebody owns.
Text recognition makes mistakes everywhere, and the expensive ones in a contract are in the numbers and the defined terms. A garbled word can be caught on a careful read. A misread digit in a rate table usually is not, because nothing around it looks wrong.
So spot-check the pages where the numbers live: fee schedules, insurance limits, notice periods, the effective date. Put the converted text next to the page image and read the figures against each other. When a figure cannot be read with any confidence, ask the other side for the native file. A counterparty who generated the document last week usually has one, and it is a much easier request to make before signature.
After conversion you are holding two versions of the file, and somebody should write down which is which. Usually the signed scan is the record and the converted file is the working copy review runs on. The answer matters less than having picked one.
That distinction earns its keep as soon as anyone compares anything. The legal blackline comparison in Word takes two documents and shows what changed between them in a third, and it needs real text on both sides. A playbook check needs the same thing, and so does a clause search. So check the conversion on the way in. Finding a bad one halfway through a review costs you the review.
Someone has to do all of this on every inbound file, and that is the part that fails quietly. At a few dozen files a month it is an annoyance somebody absorbs. At several hundred it stops happening on the busy weeks, which are the weeks the mistakes get made. Software will not tell you whether the text recognition read your liability cap correctly, though a review workflow that runs text recognition on scanned uploads will at least stop image-only files from reaching a first pass with nothing in them to read.
Keep a count of the inbound files that arrive image-only, broken out by source. It is worth keeping because of what it lets you ask for.
Some counterparties send scans every single time, because their process ends at a printer. When you can tell procurement that most of the last dozen files from a supplier arrived as photographs, there is something specific to raise at the next renewal: send the Word version. That can fix more files than anything you build on your own side.
Scans are going to keep coming in. Signature pages, agreements pulled out of a filing cabinet, anything that spent part of its life inside a fax machine. The work is catching them at intake, converting them, checking the pages with numbers on them, and handing the reviewer a file that can be searched.
goHeather is free to try if you want to see what a first pass looks like on your own inbound paper.
This is legal information, not legal advice; consult a lawyer for legal advice
Jeff Dutton is a lawyer who advises on technology, corporate, privacy, commercial, employment and real estate law.
Jeff founded his own small law firm, Dutton Law, in 2016 (and merged it with a larger firm in 2019). Before that, Jeff was a prosecutor and a commercial law lawyer at a national boutique law firm.
Jeffrey is a frequent lecturer on legal matters and has been published in newspapers and trade journals. In addition, Jeff was the editor and co-author of a leading employment law text for lawyers for many years.
Education:
Western University, BA (2009)
University of Ottawa, Faculty of Law, JD (2012)

Get the latest contract tips, updates, and exclusive content straight to your inbox. Subscribe now and never miss out on what's new in contract law or at goHeather!
Our AI sifts through each clause, identifying potential risks. This enables us to provide quick yet comprehensive contract reviews, equipping you with the legal information you need to make informed decisions.