What is OCR (optical character recognition)?

Optical character recognition lets software read printed or handwritten text from a picture. A scanned contract, a photographed receipt, or a PDF made from a scan is, to a computer, just an image. OCR finds the letters and words in it and converts them into text that can be copied, searched, edited, or passed to other software.

For example, an accounts team receiving paper invoices can scan them and use OCR to pull out the supplier name, invoice number, date, and total. An automation can then check those details against purchase orders and enter them into the accounting system, leaving a person to review only the invoices that do not match.

OCR has existed for decades, and newer AI-based versions tend to cope better with messy layouts, tables, and handwriting. Some tools go beyond reading characters to understanding documents, for instance recognizing which number on an invoice is the total and which is the tax, an approach sometimes called intelligent document processing.

OCR still makes mistakes, especially with poor scans, unusual fonts, faded print, handwriting, and characters that look alike, such as the letter O and the number zero. An error in an amount or an account number can be costly, so OCR results that feed into payments or records should be validated automatically where possible and reviewed by a person when anything looks off.

An example

A team scans signed paper forms and uses OCR so each form's text can be searched and its key fields copied into a spreadsheet.