What does OCR mean?

A scanned document is at first just an image: the computer sees dots, not letters. OCR recognises the characters in this image and outputs them as text. Only then can the document be searched, copied and evaluated.

How does OCR work?

  • Prepare the image: straighten it, increase contrast, remove noise.
  • Recognise characters: find and assign lines, words and letters.
  • Output text: for example as a searchable PDF in which the recognised text lies behind the image.

What OCR cannot do

OCR is only as good as the original. A clean scan is read better than a crumpled till receipt, printed text better than handwriting. And OCR only delivers text: it does not recognise that “4711” is the invoice number and not the house number. That takes a second step that assigns the characters to the fields.

OCR and e-invoices

An e-invoice does not need OCR. With XRechnung and ZUGFeRD the invoice data is already structured in the XML and is read directly. OCR remains for everything that arrives on paper or as a plain PDF. More on this in the guide to the e-invoicing mandate.

OCR and the GoBD

If a scanned document is enriched with full text by OCR, for example as a searchable PDF, this full text must be retained for the duration of the retention period after verification and correction and made available for audit purposes (GoBD para. 130).

Where OXS.eco uses OCR

With OXS.eye you photograph a document with your smartphone. The app creates a PDF, reads out the text by OCR and sends both to OXS.eco. In invoice receipt, a document AI reads the invoice fields and suggests them. Checking and approval are done by your team. How this works step by step is shown in the guide AI in accounting.

Frequently asked questions about OCR

What does OCR stand for?

OCR stands for optical character recognition: the image of a document is turned into machine-readable text.

How accurate is OCR?

That depends on the original: a clean scan is read better than a crumpled till receipt. That is why we do not quote a fixed rate. Uncertain fields should be checked before posting.

Do you need OCR for e-invoices?

No. With XRechnung and ZUGFeRD the data is structured in the XML and is read directly.

What is the difference between OCR and AI recognition?

OCR turns an image into text. AI recognition assigns this text to fields, for example invoice number, date and amount.

See how a document is read

In the demo we use sample data to show how OXS.eco reads an invoice, suggests the values and presents them to your team for checking.