Optical character recognition (OCR)
Also known as: OCR, text recognition
Optical character recognition (OCR) is technology that detects text in images, such as scanned documents, photos, screenshots, and PDFs, and converts it into digital text that can be searched, copied, and edited.
It's used for digitizing paper records, reading receipts in expense apps, translating signs through phone cameras, and making scanned PDFs searchable. While modern OCR, boosted by AI, is very accurate on clear printed text, handwriting, poor scans, and unusual layouts still cause errors.
Optical character recognition (OCR), explained at three levels
At home
OCR is like a fast typist looking at a photo of a page and typing out every word they see. To your computer, a scan is just a picture made of dots, so it can't search or copy the words. OCR software studies the shapes of the letters, works out what each one is, and turns the picture into real text you can search, copy and edit.
At work
Optical character recognition, or OCR, converts text in scanned documents, photos and PDFs into real digital text. Scanners and PDF tools use it to make paper records searchable and copyable, and expense apps use it to read receipts. Businesses care because it saves staff from hunting through scanned files by eye, though OCR can misread poor scans, so important numbers need checking.
At home
OCR turns a picture of text into actual text a computer can search, copy and edit. The software first cleans up the image, straightening it and boosting contrast, then finds the areas that contain text and splits them into lines and words. Modern OCR uses neural networks trained on huge numbers of examples to recognize each word, using surrounding letters for context. Phone features such as Live Text and Google Lens do this instantly on photos.
At work
Scanners and PDF tools can add an invisible text layer under the scanned image, so the PDF looks unchanged but becomes searchable and copyable. Accuracy depends on image quality: 300 DPI scans, good contrast, straight pages and clean originals work best, while handwriting, unusual fonts, stamps and tables are harder. Intelligent document processing builds on OCR to extract fields, such as invoice numbers and totals, into structured data.
At home
OCR can confidently misread similar characters: a 5 as an S, a 0 as an O, or a 1 as a lowercase l, especially on faded receipts or crumpled paper. A single wrong digit in an amount, account number or date can cause real problems if you copy it into a payment or tax form without checking. In my view, always double-check amounts and dates against the original before relying on them.
At work
Validate extracted data with confidence scores, cross-field checks such as totals matching line items, and human review for critical fields like payment amounts and bank details. Protect scanned documents, which often contain personal information, with the same controls as originals. Choose OCR engines that handle French accents and bilingual documents well for Canadian use. Check where cloud OCR and AI services process and retain data before sending confidential scans.
Real-world examples
At home
Lena photographed a recipe from her grandmother's cookbook. Live Text turned it into text she could copy into her notes app.
At work
Jerome's Saskatoon MSP set up a client's scanner to save searchable PDFs, so staff could find decades of land records by typing a name.
How it compares
| feature | Scanned image | OCR |
|---|---|---|
| Output | Picture of a page | Searchable, editable text |
| Searchable | No | Yes |
| Accessible to screen readers | No | Yes |
| Accuracy | Exact visual copy | Can contain errors |
Scans capture appearance. OCR captures the words.
For analysts
In document digitization projects, I'd define accuracy targets and review steps for critical fields. OCR errors in names or amounts can cause real problems downstream.
History & origin
Early reading machines were developed in the early 20th century, and OCR was used commercially by the 1950s and 1960s for tasks such as reading postal codes and bank cheques. Ray Kurzweil's company introduced the Kurzweil Reading Machine in 1976, which read printed text aloud for blind people. Hewlett-Packard developed the Tesseract OCR engine in the 1980s, released it as open source in 2005, and Google then led its development. Smartphones made OCR everyday technology through apps such as Google Lens, launched in 2017, and Apple's Live Text in 2021. Multimodal AI models in the 2020s extended OCR to complex documents and handwriting. In my view, OCR is the bridge that keeps paper from becoming a dead end.