Voice scam check

How to spot an AI voice clone scam

A message from your daughter on a new number: her phone was stolen, she is stuck, and money has to go today. The pattern is common enough that the US Federal Trade Commission keeps a whole category for it — family emergency scams — and voice cloning is what makes it work: if the call sounds like your relative begging for help, you are far more likely to act (FTC, family emergency scams).

What changed is the check everyone used to rely on: asking them to call. A scammer can now sound like the person you know, which means the voice alone is no longer proof of who is on the line (FTC, voice cloning).

Direct answer: five checks before you send money

1. Call back on a number you already have. Find the person in your own contacts and ring that number, not the one the message or call came from. If you cannot reach them, work through another family member or a friend (FTC).
2. Agree a family code word or question. Fix one detail that only your household knows and use it whenever money is requested. A cloned voice reproduces how someone sounds, not what only the two of you know.
3. Treat urgency and secrecy as the tell. The problem has to be solved right now, there is no time to come by, and often nobody else may be told. That pressure is the mechanism: it exists to stop you checking. Take the hour anyway.
4. Read the payment route, not the story. The FTC names the classic signs: a request to wire money, buy gift cards and read out the numbers, or send cryptocurrency — payment methods that make money hard to recover (FTC, March 2023). A genuine emergency can also be handled with a payment you can reverse.
5. Let one other person look at it first. Show the message or describe the call to another family member or to your bank before transferring anything. These scams rely on shame and speed; one second opinion usually breaks the pattern.

Why a cloned voice is so convincing

Cloning a voice does not require an interview. The FTC describes a scammer taking a short audio clip of the person — often something already posted online — and using a cloning program so the call sounds like the real thing (FTC, March 2023). The caller therefore does not need to know anything only your family would know. Sounding familiar is enough.

That also explains the shape of the call: kept short, with a reason why the person cannot be reached on their own number, because every minute you spend checking is a minute the story loses.

What a detector does and does not see

A scam request arrives as text, an image, or a call. You can run the message text through the text detector and a screenshot or profile photo through the image detector. That is triage: it helps you decide which message to verify first, and it says nothing about the sender. Short messages carry too little writing style for a reliable signal, and a genuine panicked message can pick one up.

More importantly: a recording of a voice is not evidence of identity. A text or image score says nothing about whether the speaker exists, whether the voice is synthetic, or whether an emergency is real. Our checkers do not assess live phone calls, so what you hear cannot be verified with a score.

What a cry for help cannot tell you

• whether the sender really is your relative or friend, even when the voice or the profile photo matches;
• whether the number belongs to that person — a prepaid SIM or a hijacked account looks identical;
• whether an emergency exists; only the person, or someone who can reach them, knows that;
• whether it is fraud; only a bank or an investigating authority can establish that.

Treat the request as a prompt to verify through your own contacts, never as proof that something is wrong.

If you already sent money

• Contact your bank or card issuer immediately and ask about a recall or block while that is still possible; speed decides the outcome.
• Keep everything: the message or call details, the number, the profile photo, the payment proof, the account and the timestamp.
• Report it to your national fraud reporting body and, where there is loss, file a police report. In the US, scams can be reported at ReportFraud.ftc.gov (FTC).
• Tell your family: the same number or account is often reused, and saying it out loud removes the shame that keeps these cases quiet.

More on judging this kind of evidence is in the guides on spotting a fake payment proof, deepfake investment ads with celebrities and a deepfake on a video call and fake debt collection letters. If the request leads on to a platform that shows returns, follow the five checks on a fake investment platform. To check a file or a piece of text on its own, use the free checker. If the sender's own account is being used without their knowledge, follow the five steps for a hijacked WhatsApp account.

Frequently asked questions

I heard my child's voice — how can that be fake?

Because a voice can be cloned from a few seconds of audio that already exists online. Tone and rhythm can match while the caller is someone else. The voice therefore stopped being a check; calling back on your own saved number still works.

Is a message from a new number always a scam?

No, people really do change numbers. Just never treat it as proof of identity: call the old number you have, or confirm through someone else, before any money moves.

Can I upload the screenshot or the message text?

Only imagery you are permitted to use and that contains no other people\'s personal data. Your own message is usually fine; remove names and numbers where you can.

What if the caller knows our code word?

Then treat it as one extra signal, not proof. Verify through a number you found yourself, and never share amounts, codes or payment details on the strength of one call.