News
tech

A Scanned Lease Has No Text Layer, So Your Search Finds Nothing

A scan stores the page as a picture, so the words you can read are not words the file can match. The W3C puts it plainly: the content of the document is images, not searchable text. Here is the two minute test, and why an empty search result is the part that costs you.

Two PDFs of the same lease can look identical on screen and behave nothing alike. One holds the characters of the document. The other holds a photograph of them. Which one you have was decided by a scanner setting, years ago, by someone who may no longer work there.

What is in the file

A scanner captures light, not language. It makes an image of the page, wraps it in a PDF container, and stops. The letters you read are shapes, the same way a brick wall in a photo is shapes. Optical character recognition is the separate step that reads those shapes and writes the characters into an invisible text layer under the picture. The W3C puts it without hedging in its OCR technique for WCAG: a document that consists of scanned images of text is inherently inaccessible because the content of the document is images, not searchable text. Assistive technology cannot read it either.

The empty result is the part that costs you

Search your lease folder for the word pet and get nothing back. That result has two meanings. Either no lease in the folder mentions pets, or none of your leases can be searched at all. Your computer reports both with the same blank screen and the same quiet confidence.

That is the failure worth caring about. A search that errors teaches you something. A search that returns zero teaches you something false, at the moment you are checking a fact rather than browsing. You ask whether unit 304 ever had a parking clause, the archive says nothing, and you answer the owner from memory while believing you checked.

Test it in two minutes

  1. Open one of your oldest signed leases and try to drag the cursor across a line of body text. If letters highlight, there is a text layer. If the cursor draws a rectangle over the page instead, there is not.
  2. Use the find command and search for a word you can see with your own eyes, such as rent or tenant. No match on a word that is visibly there is the whole diagnosis.
  3. Repeat on your newest lease, and on one that arrived as a phone photo. The answer often differs, because the scanner, the signing platform and the camera each make this choice on their own.

You are mapping which part of the archive is readable and which part is wallpaper.

OCR is a reading, not the record

Recognition is a guess with a confidence score, and it guesses worst on the content property managers lean on: handwritten addenda, initials in a margin, a faxed form that was already a copy of a copy, a page fed in sideways. A unit number can come through as 3O4 with a letter in the middle, unsearchable while looking fine on screen.

So treat the text layer as an index, not as the document. The picture remains the record of what was signed, and the layer underneath can be wrong without the page being wrong. That two layer structure is also why a black box drawn over text is not a redaction: you covered the picture and left the characters.

Fix the intake before the backlog

Turn recognition on where documents enter. Most scanners have a searchable PDF output setting, and Acrobat can run recognition on scans it takes in. Whatever tool you already pay for probably has the option switched off, and changing it once makes every future file findable.

Then work backwards through the backlog in the order you would search it: current leases, then notices served, then inspection reports. Rename as you go, because recognition puts words inside the file and does nothing for a file called scan0043.pdf.


Back to all articles