OCR (optical character recognition)

What is OCR?

OCR is software that recognises letters and numbers on an image and turns them into text the computer can read. OCR stands for optical character recognition. In Dutch it is called tekstherkenning.

To the computer, a scan of an invoice is just a picture. After OCR you can search the text, copy text and extract data from it.

How does OCR work?

The software searches the image for shapes that look like letters and numbers, and converts them into text. The better the scan, the better the result. Skewed photos, creases and handwriting cause more errors.

Ordinary OCR only produces text. It does not know which number is the total amount. That requires document recognition: software, often with AI, that recognises the structure of a document. It finds, for example, the supplier, the date and the VAT. Microsoft has such a service, with a ready-made model for invoices.

Example in an SME

What should you look out for?

• See what your package can already do. Some accounting packages recognise purchase invoices themselves. AFAS, for example, has Scan & herken for this. Other packages work with a separate service that you connect.
• An e-invoice does not need OCR. An e-invoice is a file that software can process directly. A PDF by email is not an e-invoice: it does need to be recognised.
• Checking remains necessary. Recognition makes mistakes. Have an employee look at what deviates or where the software is unsure.
• Privacy. If there is personal data on the documents, the GDPR applies, including at the service that does the recognition.

After recognition, an accounting integration puts the data into your administration. You can read how that works together in Reading in invoices automatically. Do you currently enter invoices or packing slips by hand? Then read Invoices and packing slips are entered by hand, or see how I approach document reading and automation.

Frequently asked questions

Sources: AFAS: reading in invoices via Scan & herken (in Dutch), Microsoft Learn: Azure Document Intelligence and Logius: e-invoicing, how it works (in Dutch). Consulted on 7 October 2026.