OCR-B Font
OCR-B is a monospaced typeface designed to be read reliably by both machines and people. Adrian Frutiger drew it in 1968 for the European Computer Manufacturers Association, and it was later standardised as ISO 1073-2. It is the font printed in the machine-readable zone of every ICAO-compliant passport and ID card.
| Designer | Adrian Frutiger, 1968 |
| Commissioned by | European Computer Manufacturers Association (ECMA) |
| Standard | ISO 1073-2 (the “Class B” OCR alphabet) |
| Spacing | Monospaced — every character occupies identical width |
| Size used in MRZs | Size I, roughly 14 point |
| Where it appears | Passport and ID card MRZs, bank giro forms, utility payment slips |
| Sibling standard | OCR-A (ISO 1073-1), the more stylised machine-first alphabet |
Why a font needed designing
Early optical character recognition was unforgiving. Scanners of the 1960s compared shapes against fixed templates, and ordinary typefaces defeated them constantly — a lowercase l, a digit 1, and a capital I are nearly identical in most fonts, and 0 versus O is worse.
The first answer was OCR-A: a font optimised entirely for machines, with blocky strokes and exaggerated shapes that no typographer would call attractive. It worked, and it was unpleasant to read.
Frutiger’s brief for OCR-B was harder. Keep the machine reliability, but make the result legible to humans. The solution was to hold every character to identical width and spacing while keeping the letterforms close to conventional European type. Ambiguous pairs are separated by deliberate details rather than distortion — the zero is narrower than the capital O, the one carries a distinct base serif.
The monospacing does something subtle and important. Because every character sits in a fixed cell, a reader knows exactly where each character begins before it tries to recognise it. Segmentation, the step where most OCR errors originate, becomes arithmetic instead of guesswork.
Why OCR-B matters for identity verification
A passport’s machine-readable zone is not simply text printed in a particular font. The font is what makes the zone parseable at all.
Fixed character cells mean a reader can locate every field by position: characters 1 to 2 are the document type, 3 to 5 the issuing state, and so on through to the check digits. There is no ambiguity about where one field ends and the next begins, which is why machine-readable zone parsing is deterministic in a way that general document reading is not.
This matters more than it sounds. General optical character recognition on the printed side of a document has to cope with arbitrary fonts, varying layouts, and design choices that differ between every issuing authority. The MRZ has none of that variability, which is precisely why check digits over MRZ data are meaningful — the character set is constrained enough that a transcription error is detectable rather than plausible.
Reading the MRZ accurately is the first step in identity document verification, and it gates what comes after. The chip cannot be opened without MRZ-derived keys, so an OCR failure at this stage blocks every cryptographic check downstream.
OCR-A vs OCR-B
| OCR-A | OCR-B | |
|---|---|---|
| Standard | ISO 1073-1 | ISO 1073-2 |
| Designed for | Machine reading above all | Machine reading with human legibility |
| Appearance | Blocky, uniform stroke weight, stylised | Close to conventional European type |
| Spacing | Monospaced | Monospaced |
| Typical use | Cheques, legacy banking documents | Passport and ID MRZs, payment slips |
ICAO chose OCR-B for travel documents for a practical reason: a border officer needs to read the same line the scanner reads, and OCR-A is genuinely hard on the eye.
What OCR-B can’t do
It carries no security properties. OCR-B is a public standard and a freely available typeface. Anyone can set text in it. Its presence on a document proves formatting compliance, nothing more.
It does not validate the data. A forged MRZ printed in perfect OCR-B parses exactly as cleanly as a genuine one. Validity comes from the check digits and from cross-matching the MRZ against the chip and the printed data page.
It only covers the MRZ. Everything else on the document — name, address, issuing authority, the visual inspection zone — is set in whatever typeface the issuer chose, and reading it requires general OCR with all the variability that implies.
Print quality still breaks it. Worn, thermally faded, or poorly printed MRZs degrade recognition regardless of the font, which is why capture quality remains the practical bottleneck in mobile verification.
Frequently asked questions
Is OCR-B free to use?
The design is published as an ISO standard and multiple free implementations exist. Individual digital font files may carry their own licences, so check the specific file rather than assuming.
Why is the MRZ monospaced?
Fixed character width lets a reader locate every field by position before recognising a single character. It converts segmentation from a hard computer vision problem into simple arithmetic, which is the main reason MRZ parsing is so much more reliable than reading the rest of the page.
What is the difference between OCR-B and MICR?
OCR-B is read optically — a camera sees the shapes. MICR characters are printed with magnetic ink and read by sensing a magnetic waveform, which is why they survive being stamped, folded, and marked over. Both were designed to solve machine reading, using entirely different physics.
Does every ID document use OCR-B?
Every ICAO-compliant machine-readable travel document does, in its MRZ. National ID cards outside that specification, and documents with no MRZ at all, may use anything.
