How to do it

The only reliable verification doesn't run through listening, but through cross-checking. Here are the steps, from the simplest to the most solid.

  1. Listen to the context, not just the timbre. Generated audio often has absent or unnatural breathing, flat intonation at emotional moments, background noise that never changes, sharp cuts between sentences. These are weak clues (systems are improving), but if you notice them, raise your guard.
  2. Break in and ask a verification question. Ask something only the real person can know and that can't be found online: a detail shared recently, a private nickname, an event you were at together. A cloned voice recites a script, it doesn't improvise personal answers.
  3. Hang up and call back yourself, at the number you have in your contacts. It's the check that dismantles almost every scam: the scammer doesn't control the person's real number. If the request was genuine, you reach it by calling back; if it was fake, you discover the person is fine and never called you.
  4. Agree in advance on a family safe word. A secret word or phrase, not obvious and not findable online, to ask for in case of an emergency call with requests for money. If the caller doesn't know it, you hang up.

Check: if, calling back at the usual number, the person answers calmly and knows nothing about the "emergency," you were dealing with a fake. The independent verification worked.

For unexpected calls asking for money or urgent data, this rule applies in the family: you always break off and call the other person back at their known number before doing anything. No real emergency prevents you from hanging up and calling back.

A concrete example

Antonio gets a call: the voice is his son's, agitated, saying he's had an accident abroad and urgently needs a transfer for the lawyer. The voice is identical. But Antonio remembers the family rule. He says "I'll call you right back" and hangs up, ignoring the protests of the caller who insisted on staying on the line. He dials his son's number from his contacts: he answers calmly, he's at the office, nothing happened. The voice on the phone was a clone built from some public video. Antonio didn't lose a cent because he verified independently, instead of trusting his ear.

When it does NOT work (and how to fix it)

If the voice is indistinguishable from the real one

A few seconds of audio are enough today to generate a convincing clone, and a sample of a few minutes produces a voice that most people can't tell apart over the phone. When your ear doesn't help you, stop relying on it: the only defense is independent confirmation (calling back, the safe word, the personal question). Verification isn't done with your ears, it's done over a second channel.

If they tell you not to hang up and not to call anyone

That's exactly where the scam hides. The pressure to stay on the line, the secrecy ("don't tell your mother"), the isolation from other people are tactics that real emergencies don't require. The remedy is to flip the rule: precisely because they tell you not to, hang up and verify. A legitimate request survives a verification call.

If they ask for payments in unusual forms

Gift cards, cryptocurrencies, instant transfers to strangers, peer-to-peer payment apps: these are the favorite tracks of scams because they're hard to trace and to reverse. No real lawyer, hospital or public agency asks to be paid in gift cards. Faced with a request like that, consider the call fake until proven otherwise and verify by another route.

A tip from someone who actually uses it

Decide on the family safe word today, with a clear head, not when the phone rings. Choose it together with elderly parents, children, your partner: a strange word no one would find on social media. And add the callback rule: faced with any urgent request for money, you hang up and call back. They're two free habits worth more than any detection software.

Frequently asked questions

Is there an app that recognizes fake voices during a call?

Detection tools are emerging, but none is reliable enough to trust during a real phone call, and clones improve fast. Don't wait for technology: the defense that works right away is behavioral, that is, calling back and using the safe word.

Is it enough not to post my voice to be safe?

Reducing public audio helps, but it doesn't eliminate the risk: sometimes a few seconds from a voice message or a video are enough. And in any case you might be the one receiving the call with someone else's cloned voice. That's why the defense isn't only protecting your own voice, but having a verification procedure that applies to every suspicious call.