The Scanned Contract Problem: How to Make a Photographed Agreement Searchable

Someone signs a contract, photographs it with their phone, and emails you the PDF. It looks like a document. Open it, press Ctrl+F, search for “termination”, and nothing is found — because there is no text in the file at all. There is a picture of text, and pictures are not searchable.

This matters more than it sounds. A scanned contract cannot be searched, cannot be quoted without retyping, cannot be analysed by anything, and cannot be read by a screen reader. It is, in every practical sense, a photograph you are storing in a legal file.

---

Telling a scan from a document

Open the PDF and try to select a sentence with your cursor.

• A blue selection appears over the words — there is a text layer. Nothing to do.
• A blue box appears over the whole page, or nothing happens — it is a scan.

A file can be both: a contract exported from Word with a photographed signature page appended is text for nine pages and a picture for one. That last page is usually the one with the signatures and the date on it.

---

Fixing it

Make PDF searchable — upload the scan, get the same PDF back with an invisible text layer added underneath the image. The pages look identical; the difference is that Ctrl+F now works, text selects and copies, and anything downstream can read it.

This is the version to keep. It preserves the original appearance — including the signatures — while making the content reachable.

Extract the text if you want the words rather than the document, and image to text if what you have is a JPG or PNG rather than a PDF.

Up to 30 pages per document, no account.

---

What OCR gets wrong

Text recognition is very good and not perfect, and the errors cluster in the places that matter in a contract:

Numbers. 1 and l, 0 and O, 5 and S. A payment of $1,500 read as $l,500 is not searchable and not obviously wrong at a glance.

Handwriting. Dates written by hand in a blank, initials in a margin, anything added after printing. Treat handwritten fields as unread.

Bad scans. A photograph taken at an angle, in poor light, or of a photocopy-of-a-photocopy. Every generation of copying costs accuracy.

Stamps and signatures over text. Where ink overlaps a printed word, one of them wins and it is usually the ink.

The practical rule: OCR is reliable enough to search and to analyse, and not reliable enough to quote from without checking the original page. Before you paste a figure into an email, look at the scan.

---

Getting a better scan in the first place

Most OCR failures are input failures, and the fix costs nothing:

• Flat surface, no shadow across the page, phone parallel to the paper rather than at an angle.
• Scan at 300 DPI if you have a scanner. Below 200 the error rate climbs sharply; above 400 you get a larger file and no more accuracy.
• Black and white text on white paper — colour adds size and nothing else.
• One page per image. A two-page spread photographed together halves the effective resolution of each page.

If the file is now too large to send, compress it afterwards rather than scanning at a lower quality.

---

Then read it

Once the text layer is there, the contract is a contract again. You can upload it for a free review — six scores out of ten with the reasoning for each, and the number of concrete weaknesses found — the same as for any other document.

That is the actual point of the exercise. A scan sitting in a folder is a liability nobody has read; a searchable one is a contract you can check.

---

Next step: make your scan searchable → — free, no account. Then see what it scores.