What is OCR and how does it work?
Note: OCR results can contain errors. Check important names, dates, amounts and instructions against the original document.
Optical character recognition (OCR) detects text in an image or scanned page and produces text that software can search, copy or process. It is useful, but its output is a recognition result—not a guaranteed transcription or an understanding of the document.
What OCR does
A scanned page may look like text to a person while a computer sees only pixels. OCR analyzes the image, locates text and predicts characters and words. Depending on the software, it may also return line positions or parts of the page layout.
OCR is different from simply opening a digital PDF: many PDFs already contain selectable text. OCR is mainly useful when the page is a photo or image without a text layer.
How recognition happens
A typical workflow prepares the image, identifies text regions, recognizes characters or words, then arranges the result into lines or other structures. Some modern services can also detect layout elements or multiple languages, but supported languages and features vary by product and version.
The output may be plain text, a searchable PDF, or structured data. Standard OCR does not by itself confirm that an extracted amount, name or sentence is factually correct.
How to improve a scan
Use a sharp, evenly lit image with the page in focus and as little skew as practical. Tesseract’s documentation says the engine works best at 300 dpi or above; treat that as guidance for that OCR engine, not a universal requirement for every device or service. Follow the receiving organization’s specifications.
Keep the original and review the output. Check numbers, dates, names, accents, columns, tables, handwriting and faint notes. A recognition tool can confuse similar characters or reorder complex layouts.
Privacy and document handling
Before uploading a document to an online OCR service, check the provider’s current privacy, retention and access terms. A page can include personal or financial information. If the document is sensitive, consider an approved local workflow and follow the rules of the organization that owns it.
ConverteArq’s PDF tools do not currently provide OCR. You can learn how to prepare phone scans in our related guide; this article does not promise text extraction by the site.
Related guide
For a clearer source image, read how to scan documents with a phone.
Frequently asked questions
Does OCR understand a document?
OCR recognizes text and may return layout information. That does not mean it verifies facts, understands every relationship or reliably interprets legal or financial meaning.
Is 300 dpi required for every OCR tool?
No universal setting applies to every tool or document. Tesseract recommends at least 300 dpi for its engine; the receiving organization and the specific software may have different requirements.
Can OCR read handwriting?
Some products support handwriting, but capability varies by language, writing style, image quality and model. Review the result against the original.