Scanned PDF invoice into XRechnung/ZUGFeRD? What OCR can and cannot do

„Can I simply convert my scanned invoices into XRechnung?“ – this question comes up often. The answer is a cautious yes, with clear limits.

> Quick answer: A scanned PDF is an image without structured data and is therefore not an e-invoice. With AI/OCR, the details can be extracted and transferred into an XRechnung or ZUGFeRD file – but with pure scans, recognition is more error-prone, which is why checking the data is essential.

A scan is not an e-invoice

An e-invoice under EN 16931 consists of structured data (XML). A scan, by contrast, contains only an image of the invoice – to a computer these are at first just pixels, not readable fields. Only conversion into a structured format turns it into an e-invoice.

What AI/OCR can do

Modern, AI-supported extraction can read invoice data even from image PDFs: invoice number, date, line items, amounts, tax, seller, and buyer. An XRechnung or ZUGFeRD file can then be generated from this data.

Where the limits lie

Reliability depends heavily on the source:

• Digitally generated PDFs (exported from a system) can be read most reliably.
• Pure scans – especially skewed, low-contrast, or low-resolution ones – lead to recognition errors more often.

So the rule is: do not rely blindly on recognition. A check of the extracted data before generation is mandatory – which is exactly what a review step is for.

Recommendation

Wherever possible, generate the invoice digitally in the first place and export it as a PDF. Scanning remains the fallback for paper documents – and even then you should verify the data carefully.

Conclusion

You can create an e-invoice from a scan with AI/OCR, but the quality of the source is decisive. Always check the recognised data before you generate the XRechnung or ZUGFeRD file.

---

TaxLayer extracts invoice data with AI support and lets you check every detail before generation. Questions to info@landauer.de.