A wrong ABN and a smudged one look the same to OCR. We made sure ours doesn't guess.

Every Australian Business Number carries a checksum — a modulus-89 formula baked into the digits themselves, so a wrong ABN is usually mathematically detectable, not just “looks a bit off.” Which raises an uncomfortable question for any tool that reads receipts: when the number printed on the paper fails that check, what does your extractor actually do with it?

Our early test fixtures only covered two cases — a valid ABN, and no ABN at all. Neither one exercises the case that actually matters: an ABN that’s present but checksum-invalid, the kind of thing a genuine OCR misread or a real printing error produces all the time. So we added a fourth fixture built specifically to fail the check — one digit off a known-valid number — and ran it against the live API. The extractor transcribed the ABN exactly as printed and let checksum validation flag it downstream, rather than silently rounding it to the nearest valid-looking number. 4 out of 4 fixtures passed, including the one designed to break.

That distinction matters more than it sounds. A tool that “corrects” an invalid ABN to something plausible isn’t being helpful — it’s replacing what’s actually on your receipt with a guess, and you’d have no way to tell the difference. Ours transcribes first, validates second, and never quietly substitutes one for the other.

Try Receipt Extraction →

This reflects publicly documented ABN checksum rules and is provided as software output, not tax advice.