Drop a Factur-X PDF, Skip the OCR

If a supplier in France or Germany has sent you an invoice in the last year, there is a decent chance it was a Factur-X (the French name) or ZUGFeRD (the German name) PDF. From the outside it looks like an ordinary invoice PDF. Inside, the same invoice is also encoded as structured XML, attached to the file.

Financica now reads that embedded XML directly. No OCR, no AI guesswork, exact numbers.

What's a hybrid PDF?

UBL invoices over Peppol are pure XML: machine-readable, but a human can't open one in a PDF viewer and skim it. Factur-X and ZUGFeRD split the difference. The file is a normal PDF (your supplier prints and emails it, you open it, you read it), but it carries a copy of the same invoice as a CII XML attachment glued to the back.

The format is becoming the standard for cross-border e-invoicing in the EU. France's e-invoicing mandate accepts it. Germany's accepts it (with the XRechnung variant). It is what happens when "we should send structured invoices" meets "but accountants still want a PDF they can look at."

What changes when you upload one

Until now, dropping one of these PDFs into Financica ran the same OCR/AI pipeline as any other PDF. We got the numbers more or less right, but rounding, line item ordering, and tax breakdowns occasionally drifted from what the supplier actually sent. Now:

• We detect the embedded factur-x.xml attachment at upload time.
• We read it directly. No OCR. No LLM.
• The invoice fields, line items, and tax breakdown match the source exactly. The invoice goes read-only, because the supplier's XML is the authoritative copy.

The same change applies to ZUGFeRD-tagged PDFs that arrive over Peppol or that you upload manually.

How you can tell

A green "E-invoice (Factur-X)" badge appears on the invoice header next to the status pills. Hovering it shows the profile (urn:factur-x.eu:1p0:basic, basicwl, comfort etc.) so you can tell at a glance how much detail the supplier included.

UBL invoices get the same treatment: a matching "E-invoice (UBL)" pill, so you can see structured provenance on both formats from the list view down.

The embedded XML is available too

Under Files, the attached factur-x.xml shows up as its own downloadable file alongside the PDF. If your accountant or auditor asks for the structured source, it's one click away. The Advanced tab now also surfaces the parsed profile, document type code, and business process so you can verify what was on the wire.

Supported profiles

The four common Factur-X profiles all work: MINIMUM, BASIC WL, BASIC, and EN 16931 / COMFORT. EXTENDED is read with the same parser as COMFORT. If a PDF carries a profile we don't recognize (the XRechnung CII variant, for instance), Financica falls back to the regular OCR/AI extraction so nothing breaks.

Available now. If you have a Factur-X or ZUGFeRD PDF sitting in your inbox, drag it onto the expenses page and you should see the green badge appear after a second.