How to Spot Deepfakes and Voice-Clone Scams Before They Fool You

A deepfake is synthetic media, video, audio or images generated or altered by AI to show someone saying or doing something they never did. A few years ago, most fakes had obvious glitches. Today, a convincing clip can be produced with consumer software, and a cloned voice can be built from a short recording posted online. The practical lesson is that you can no longer rely on your eyes and ears alone. You need habits that work even when the fake is perfect.

The scams to know

The family emergency call

The Federal Trade Commission has warned that scammers use voice cloning to make the old "grandparent scam" more believable. You get a call from what sounds like your grandchild, child or spouse in distress: an accident, an arrest, a kidnapping. There is urgency, a request for secrecy, and a demand for money through wire transfer, gift cards, cryptocurrency or a courier picking up cash.

The fake boss

Businesses have been targeted by impersonation of executives, through cloned voices on phone calls or, in some reported cases, video calls featuring fake participants. The request is typically an urgent payment or a change in bank details.

Celebrity and investment deepfakes

Fake videos of well-known figures endorsing investment schemes, crypto giveaways or miracle products circulate as social media ads. A familiar face does not mean a real endorsement.

Romance and sextortion

Scammers use AI-generated photos to build fake profiles, and some create sexualized fakes of real people to extort them. Teenagers have been particular targets.

Political and news fakes

Fabricated audio or video of public figures can spread quickly, especially around elections and breaking news. In 2024, the Federal Communications Commission ruled that AI-generated voices in robocalls count as "artificial" under the Telephone Consumer Protection Act, making such robocalls illegal without prior consent.

Visual and audio clues (useful, but fading)

Generation quality improves constantly, so treat these as reasons for suspicion, not proof either way:

• Faces: odd blinking, mismatched lighting on the face versus the background, blurring where the face meets hair or neck, teeth or ears that look smeared.
• Hands and text: extra or fused fingers, garbled lettering on signs or badges, though newer models handle both better.
• Lip sync: mouth movements that drift from the audio, especially on sounds like "b," "m" and "p."
• Audio: flat emotion, unnatural pacing, missing breaths, or a voice that sounds right but uses phrases the person never would.
• Physics: shadows, reflections and earrings that do not behave consistently.

What still works: verify the context

The most reliable checks do not depend on spotting pixels.

Protecting your family from voice-clone calls

1. Agree on a family code word that only close family knows, and that would never be posted online. Ask for it in any emergency call asking for money.
2. Hang up and call back on a number you already have for that person. Scammers can spoof caller ID, so an incoming number that looks right is not proof.
3. Ask a question only the real person would know, and avoid details that appear on social media.
4. Never pay by gift card, wire, crypto or cash courier in response to an unexpected call. Those payment methods are hallmarks of fraud.
5. Talk to older relatives about this scam in advance. People are far less likely to fall for it if they have heard of it before.
6. Consider your public audio. Long videos of your voice on public social accounts make cloning easier. Tighten privacy settings where practical.

Protecting a workplace

• Require callback verification for payment requests or bank detail changes, using a known number from internal records.
• Use two-person approval for large or unusual transfers.
• Train staff that a familiar voice or face on a call is not authentication.
• Set up a clear reporting path so employees can flag suspicious requests without fear of embarrassing a senior colleague.

If you have already been fooled

Scams succeed because they are designed to, and people of every age and background fall for them. If you sent money or information, act quickly and without embarrassment:

1. Call your bank, card issuer or payment app immediately. Ask whether the payment can be stopped or reversed. For wire transfers, every hour matters.
2. If you paid with gift cards, contact the card issuer right away with the card numbers and receipts; some can freeze unused balances.
3. Change passwords and turn on two-factor authentication if you shared any login details.
4. Freeze your credit with the three nationwide credit bureaus if you shared your Social Security number. A freeze is free.
5. Report it (see below). Reports help investigators connect cases and warn others.
6. Beware of "recovery" scams. Fraudsters often contact victims again offering to retrieve lost money for a fee.

The legal picture

US law on deepfakes is evolving but not empty:

• Fraud using a deepfake is still fraud, covered by existing federal and state laws.
• The FTC's impersonation rule targets scammers who impersonate government agencies and businesses.
• The TAKE IT DOWN Act, signed into federal law in 2025, criminalizes knowingly publishing non-consensual intimate images, including AI-generated ones, and requires covered platforms to remove reported images within a set time after a valid request.
• Many states have passed laws addressing sexual deepfakes, election-related deepfakes or unauthorized use of a person's voice and likeness.

Where to report

• Fraud and scams: the FTC at ReportFraud.ftc.gov and the FBI's Internet Crime Complaint Center.
• Money already sent: contact your bank or payment provider immediately; speed matters.
• Intimate images: report to the platform, and use the National Center for Missing & Exploited Children's Take It Down service if the images involve a minor.
• Impersonation on social media: use the platform's impersonation reporting tools.
• Local police: for threats, extortion or in-person pickups.

The mindset shift

The goal is not to become a forensic expert. It is to stop treating a voice, a face or a video as proof of identity or truth on its own. Pause when content triggers strong emotion, verify through a channel you control, and check where something came from before you share it. Those habits work no matter how good the fakes become.