Checking incoming e-invoices: what the BMF asks for, and what it does not
Since the BMF letter of 15 October 2025 there are three named classes of error — and only two of them touch the input VAT deduction. This page sorts out what a technical check on receipt can do, and what it explicitly cannot.
The three classes of error
The letter separates them because they have different legal consequences. That is the change that matters most in practice: "the invoice is faulty" has not been a useful sentence since.
1 · Format errors
The file does not meet the EN 16931 syntax — wrong file format, broken XML, missing technical mandatory fields. Consequence: it is not an e-invoice but an "other" invoice, and once the transition periods end that can affect the input VAT deduction.
Machine-detectable. This is exactly what schema validation checks.
2 · Business-rule errors
Breaches of the EN 16931 check rules and their national extension: a missing buyer reference, amounts that contradict each other. They matter for VAT only insofar as they affect a mandatory particular.
Machine-detectable — these are the BR rule codes, checked against KoSIT rule set v2026-08-31.
3 · Content errors
A wrong tax rate, an inaccurate description of the supply, an incorrect VAT ID. These bear directly on the input VAT deduction — regardless of whether the file passes every technical check.
Not machine-detectable. No validator anywhere can see that 7 % should have been 19 %.
The sentence this page will not say
"The invoice validated, so the input VAT deduction is safe." That is precisely not what the letter says. Class 3 survives every technical check, and reviewing the substance of an invoice stays exactly where it was: with a person who knows what was ordered.
What a check on receipt does is narrower and still worth having: it sorts out the two machine-detectable classes before anybody spends time on them, and it produces a dated record that the check happened.
What has to be kept
The structured part, unchanged. For VAT purposes at least the XML has to be kept as it arrived — not a PDF rendered from it and not a database extract.
The validation report. The letter names retaining validation reports explicitly; it is the record that the diligence was applied whose result you then rely on.
NormAPI stores neither the document nor the report. Both belong in your archive rather than ours — the API response is the report, and what you do with it stays with you.
Automated on receipt
The same call as on the sending side, at a different point in the process: the file as it arrived, against the official rule set.
Back comes the format and scenario that were recognised, every finding with its rule code, severity and location, and the rule-set version it was checked against. That version belongs in the report: a result without the version behind it cannot be interpreted two years later.
One case that turns up in practice and is easy to misread: when no scenario matches the profile that arrived — a ZUGFeRD MINIMUM or BASIC WL, say, which is not an e-invoice at all — the business rules are not evaluated. The report then does not say "no errors" but "not checked". We return that as its own field so that a receiving process cannot book it as a pass. For Factur-X BASIC and EXTENDED, which the official KoSIT configuration ships no scenario for, we added one: they are checked against the EN 16931 core rather than passing through unchecked.
The data, not only the file
After the check, bookkeeping wants the invoice data. The same service reads the file — XRechnung in UBL or CII, a ZUGFeRD or Factur-X PDF — into JSON: parties, lines and payment details in the same shape our generation accepts, plus the totals and VAT breakdown exactly as the invoice states them. Nothing is recomputed; what gets booked is what the invoice says. Whatever the shape does not carry is listed separately instead of disappearing quietly.
With validate=true the report and the data come in one response. Free like validation, with no allowance, and nothing is stored here either. Fields and examples are in the API documentation.
Who this is for
Tax practices that take in documents for many clients and have to sort the inbox before anything is posted.
Bookkeeping and document-management vendors that want checking and reading inside their product rather than left to their customers.
Companies running their own invoice intake that want the report filed in an auditable way.