What happens if the MRZ doesn't match the visual zone data?

A mismatch between MRZ data and visual zone data is one of the more important signals a document scanning system can surface, and how it's handled says a lot about how seriously a given solution takes fraud prevention. In a well-designed process, this kind of discrepancy doesn't get silently resolved in favor of one source or the other.

It gets flagged, because a genuine mismatch is exactly the kind of thing that indicates either a scanning error or, more seriously, a tampered document.

There are a few different reasons a mismatch might occur. The most benign is a scanning or OCR error on one side, perhaps the visual zone OCR misread a character in a name due to an unusual font or a bit of glare, while the MRZ read correctly, or vice versa.

This kind of mismatch is usually resolvable by re-scanning the document or, if the system supports it, examining the confidence scores from each extraction to determine which source is more likely correct.

A more concerning possibility is that the document has actually been altered. For example, if someone has changed the printed name or date of birth in the visual zone without correspondingly updating the MRZ, or the reverse, the two data sources will disagree in a way that a simple re-scan won't resolve, because the underlying document itself contains inconsistent data.

This is precisely the kind of discrepancy that cross-referencing MRZ against the visual zone is designed to catch, since altering both consistently, along with recalculating the correct MRZ check digits, is a meaningfully higher bar for a forger to clear than altering just one.

Because of this, most identity verification workflows treat an MRZ-to-visual-zone mismatch as a trigger for manual review rather than either accepting the document or rejecting it outright automatically, since a human reviewer can look at the actual document and images to determine whether it's a benign scanning glitch or a genuine red flag.

ScanDoc's cross-validation logic works this way by design. MRZ and visual zone data are extracted independently and then compared field by field, with any mismatch surfaced to the business rather than resolved silently, so a human decision-maker gets visibility into exactly which fields disagreed and can decide how to proceed.