How to Spot an AI Deepfake Scam Call
AI deepfake scam calls pair a cloned voice with a manufactured emergency to bypass your skepticism, and this guide covers the specific tells and verification steps for grandparent emergency scams and CEO wire-fraud calls.
An AI deepfake scam call uses a cloned voice, built from as little as three to ten seconds of audio, wrapped around a fabricated emergency designed to get you moving before you start asking questions. The voice on the line might be your daughter saying she's been in an accident, or your company's CFO saying a payment has to go out in the next twenty minutes. The cloning technology is good enough to pass a casual listen. What actually does the work is urgency plus a voice you trust, which is exactly the combination that shuts down normal skepticism.
This post is narrowly about phone calls that use a cloned voice, specifically the grandparent emergency scam and CEO or vendor fraud calls that push urgent wire transfers. It does not cover video calls or manipulated footage; for that, see the companion post on how to tell if an AI video is a deepfake. For a wider view of AI-driven scams beyond voice calls, read the broader guide to how to spot an AI scam. For the specific verification protocol to use when a caller's voice sounds right but something feels off, see the guide to spotting an AI voice cloning scam.
Why voice cloning scam calls work
Cloning a voice used to require hours of clean recordings. Now a short clip pulled from a voicemail greeting, a social media video, or a podcast appearance is enough to generate a passable copy. The tools that do this are cheap, widely available, and don't require any technical skill to operate. That's the uncomfortable part: almost anyone with a public voice sample, which is most people, is a viable target.
But the audio is only half the attack. The other half is timing. Scammers call late at night, or during a workday when a business owner is distracted, and they attach the voice to a scenario that demands an immediate decision: an arrest, an accident, a missed wire deadline. Fear and urgency narrow attention onto the crisis and away from the small inconsistencies that would normally raise a flag. For a wider look at how these tactics fit into the bigger picture of AI-enabled fraud, see the guide to AI risks.
The grandparent version of this scam typically opens with a panicked or crying voice claiming to be a family member in trouble, followed quickly by a second caller posing as a lawyer, bail bondsman, or police officer who explains how to pay. The FTC has documented this exact pattern in its consumer alert on scammers using AI to enhance family emergency schemes, warning that the cloned voice is often paired with pressure to pay through gift cards, wire transfers, or cash pickup before anyone can verify what's actually happening.
The workplace version follows the same three-part shape: a trusted voice, a request that can't wait, and a payment method that skips the usual checks. Recognizing that shape matters more than trying to judge audio quality on the fly, since by the time you're asking yourself is this call AI generated, the scammer has already gotten you halfway to sending money.
The signals that give it away
Look for these signals together, not in isolation. Any single one on its own could have an innocent explanation, but two or three at once is a strong warning:
Urgency paired with an unusual payment method: gift cards, cryptocurrency, wire transfers, or a courier picking up cash are not how real emergencies or real invoices get settled.
A request to keep the call secret, especially don't tell mom or don't loop in finance, which is designed to stop you from checking with anyone who could catch the fraud.
The caller avoids specifics a real family member or colleague would know without prompting, like a pet's name, a recent shared event, or a coworker's name, and gets vague or changes the subject when asked.
Background sound that doesn't match the story, like a supposedly noisy jail or accident scene that's actually dead quiet, or audio that cuts oddly between sentences.
A voice that sounds slightly flat, oddly paced, or has strange breathing patterns under stress, since emotional speech is one of the harder things to clone convincingly.
The caller gets defensive, raises their voice, or hangs up when you ask to verify their identity or say you'll call back on a known number.
What to actually do
If a call like this comes in, whether it's a family member or a colleague on the line, work through these steps before anything else:
Hang up and call the person back on a number you already have saved, not one the caller gives you, then confirm the story directly.
Agree on a family verification phrase in advance, a word or short phrase nobody outside the family would know, and use it whenever money or an emergency comes up on a call.
Treat any request for gift cards, crypto, or a rushed wire transfer as a signal to slow down, regardless of how urgent or upsetting the call sounds.
Verify unusual payment requests through a second channel, like a text to the person's known number or a call to their office line, before moving any money.
Never act on urgency alone. A real emergency can survive a five-minute pause to verify; a scam cannot.
If you run a business
The business version of this is CEO fraud or vendor impersonation: a cloned voice of an executive, sometimes paired with a spoofed caller ID, instructing an employee to push through an urgent wire or change a vendor's bank details. It works the same way the family version does, by combining a familiar voice with a deadline that discourages questions. A short recording of a CEO from an earnings call, a webinar, or a conference talk is often all an attacker needs.
The fix is a simple callback-verification policy: any request to move money or change payment details, no matter who it appears to come from or how urgent it sounds, gets confirmed through a second channel using a phone number pulled from internal records, not one supplied on the call. Put this policy in writing, tell everyone who touches payments about it, and make it part of the company's AI incident response plan, so there's already an agreed process if a call like this comes in.
FAQ
How much audio does it take to clone someone's voice?
Current voice cloning tools can produce a usable copy from as little as three to ten seconds of clear audio. A voicemail greeting, a video posted online, or a few seconds from a phone call is enough starting material, which is why so many people are viable targets without realizing it.
Can voice cloning scams be reported?
Yes. Report the call to the FTC, to your local police department, and to your bank or wire service if any money was sent or attempted. If the call targeted a business, report it to the FBI's Internet Crime Complaint Center as well, since these cases often get investigated together with similar reports.
Are these calls illegal?
Yes. Impersonating someone to obtain money through fraud is illegal regardless of whether a real or cloned voice is used, and several states have added specific penalties for using AI-generated voice or video to commit fraud. Illegality doesn't make the calls stoppable at the source, which is why verification habits matter more than hoping the law catches up.
How do I set up a family verification phrase?
Pick a short word or phrase with family members in person or on a call you initiated, not something that shows up in social media posts or public records. Use it any time someone calls asking for money or claiming to be in danger, and treat a refusal or confusion about the phrase as a clear sign the call isn't who it claims to be.
How did this land?
About the author
Senior Editor, AI & Product
Cecilia leads the Swarmz editorial desk. She has spent a decade turning complex AI and product topics into writing people actually finish, and she owns the blog's quality bar.
