Verification codes by voice
Voice is the fallback channel for one-time codes, which means it is used by the people for whom everything else has already failed. It has to work first time.
What goes wrong
A voice OTP is a pure digit string with no redundancy. There is no context to recover a misheard character from, no sentence around it to disambiguate, and the listener is usually typing while listening. Every weakness in digit handling shows up here undisguised.
The second failure is speed. A code read at conversational pace is faster than a person can transfer it to a keyboard, so they miss the last two digits and wait for a repeat that never comes.
Scripts you can use
Written to be read aloud by a machine, not by a person. The note under each one says which part a general-purpose voice model gets wrong.
Six-digit code
Your verification code is 4 9 2 7 3 1. Again, that is 4 9 2 7 3 1. This code expires in ten minutes.
Say it twice, always. The repeat costs four seconds and removes the entire class of "I missed the end" failures.
Alphanumeric code
Your code is B 4 K 9 2 M. That is B for Bravo, 4, K for Kilo, 9, 2, M for Mike.
B and V, M and N are indistinguishable over a compressed phone line. The second pass with a phonetic alphabet is the only reliable fix.
No code requested
If you did not request this code, do not enter it. Someone may be trying to access your account. Contact us on the number printed on your statement.
Never read out a support number the caller has not seen elsewhere — that is the exact shape of a vishing call, and training people to trust it is a security problem you are creating.
Before you record it
- Every code is read twice, at a slower pace than the surrounding sentences.
- Alphanumeric codes get a phonetic pass on the repeat.
- The expiry is stated after the code, not before it.
- The message never asks the recipient to read a code back, to anyone, for any reason.
Questions
How fast should a one-time code be read out on a call?
Slower than speech, with a beat between digits, and repeated once in full. The listener is transcribing in real time and cannot rewind; the repeat is what makes the channel reliable.
Is voice OTP delivery still worth supporting?
It is the accessible fallback when SMS does not arrive, on landlines, and for people who cannot read a screen. It is a small share of volume and a large share of the people who would otherwise be locked out.
Generate these with IVR Voice Gen
IVR Voice Gen is text-to-speech built for phone systems: digit strings, references, dates and addresses are read the way a caller can write them down, in English, French, German, Spanish and Portuguese. It is in early access.