A video call from a relative asking for money urgently. A clip of a well-known person announcing something shocking. A voice message from your boss asking you to transfer funds right now. Any of these can now be faked convincingly with AI tools that copy a face and a voice from a few seconds of real footage. These are called deepfakes, and the goal is almost always the same: to make you act quickly, before you stop to check. This guide explains how deepfake scams work, the signs that give them away, and the verification habits that protect you and your family — without needing any technical skill.
How a Deepfake Scam Actually Works
Most deepfake scams follow a simple script. First, the scammer collects public material: a video clip from social media, a voice recording from an interview, or a photo from a profile. AI tools then build a convincing copy of that person's face or voice. Next, the fake is delivered through a channel you already trust — a video call, a WhatsApp voice note, or a social media clip shared by a familiar-looking account. Finally, the message creates urgency: money is needed now, an account will be closed, a secret must not be shared with anyone else. That urgency is the real weapon. The fake face or voice only gets your attention; the rush to act is what makes the scam succeed.
Common versions include fake video calls from family members asking for an emergency transfer, fake investment videos in which a famous person appears to endorse a scheme, fake voice messages from a manager or colleague asking for a payment, and fake clips of politicians or officials announcing false news. If a message asks for money, codes, or secrecy, and it arrived out of the blue, treat it as suspicious no matter whose face or voice it carries.
Signs That Give a Fake Away
Deepfakes are getting better, but they still leave clues, especially in live calls and low-quality clips. Look for these:
- Unnatural movement: the mouth does not quite match the words, blinking looks rare or mechanical, or the head turns in a way that distorts the face at the edges.
- Lighting and skin that do not match: the face looks smoother, brighter or flatter than the rest of the scene, or shadows on the face do not match shadows elsewhere in the frame.
- Strange audio: the voice sounds flat, robotic in places, or oddly emotionless during an "emergency". Pauses fall in the wrong places, and background sounds disappear or change abruptly.
- Refusal to interact naturally: on a live video call, a scammer's fake cannot easily react to an unexpected request. If you ask the caller to turn their head, hold up a hand, or read a random word you choose, a fake often freezes, glitches, or refuses.
- Urgency plus secrecy: "Do not tell anyone", "Send it now or it will be too late", "Keep this between us". Legitimate emergencies from real family members do not normally come with a ban on checking.
- A channel that does not fit: your bank, your employer or a government office will not ask for money or passwords through a social media video or a voice note. If the channel is wrong, the message is wrong.
No single clue proves a fake, and a real video can look odd because of a bad connection. That is why the signs above should trigger verification, not a final judgement on their own. Verification is the step that actually protects you.
The 3-Step Verification Rule
Whenever a video, call or message asks you to send money, share a code, or act urgently, stop and verify through a different channel:
- Hang up or pause. End the call or stop replying. Do not argue with the caller and do not send anything while the pressure is on. A real emergency will still be real in ten minutes.
- Contact the person directly, yourself. Call them on the number you already have saved — not a number given in the suspicious message. Ask a question only the real person could answer, or use a family code word you agreed in advance. If you cannot reach them, wait. Waiting is always safer than sending money to a stranger.
- Check the claim independently. For investment endorsements and shocking news clips, search for the same story on a trusted news source or the person's official account. A genuine announcement appears in more than one place; a scam usually exists only in the clip that was sent to you.
Make this a household rule, not just a personal habit. Agree a simple family code word with parents, children and elderly relatives — a word or phrase that a scammer cannot guess from public posts. If a caller claiming to be family cannot give the code word, the call ends, no matter how real the voice sounds. Our AI voice clone scams guide explains voice-only versions of this attack in more detail.
What to Do If You Already Sent Money or Information
Act quickly, but in the right order. First, contact your bank and ask them to freeze or recall the transaction — the sooner you report it, the better the chance of recovery. Second, change the password of any account whose details you shared, from a trusted device, and turn on two-factor authentication (see our 2FA guide). Third, preserve the evidence: save the message, the number, the account details and screenshots before deleting anything, because you will need them for a report. Fourth, report the scam on the platform where it appeared, and to the authorities using the steps in our guide to reporting cybercrime. Reporting helps others too: the same fake clip is usually sent to many people at once.
Do not be embarrassed into silence. Deepfake scams are designed to fool careful people, and scammers rely on victims feeling too ashamed to report. Reporting quickly protects your money and warns the next target.
Protecting Your Own Face and Voice
You cannot stop AI tools from existing, but you can make it harder for scammers to build a convincing copy of you:
- Limit public video and audio of yourself and your children. Short public clips are the raw material for voice and face copies. Keep family videos in private groups rather than public posts where possible.
- Review who can see your posts. Check the privacy settings on your social accounts (see our social media privacy guide) and remove old public clips you no longer need.
- Be careful with voice notes in public groups. A clear voice recording is enough to clone a voice. In large or public groups, prefer text for sensitive matters.
- Talk to older relatives. The people most targeted by fake family calls are often those least familiar with AI fakes. Explain, in plain words, that a familiar face and voice on a screen no longer prove who is calling — and agree the family code word together.
Deepfakes, News and Elections: A Calmer Way to Judge a Clip
Not every deepfake asks for money. Some are made to spread false news, damage a reputation, or stir anger. Before sharing a shocking clip, ask three questions: Where did this first appear — an official account, or an unknown account that only posts such clips? Does any trusted source report the same event? And does the clip benefit someone if you share it in anger? If the answers are unclear, do not share it. Sharing a fake, even with a warning, still spreads it, because many people will only see the clip itself. The safest response to an unverified shocking video is to leave it alone and wait for a trusted source to confirm or deny it.
Frequently Asked Questions
Can a deepfake be detected by an app?
Some tools claim to detect AI-generated media, but none are reliable enough to trust on their own, and scammers test their fakes against them. Treat detection tools as a hint at most. Verification through a separate channel — calling the real person yourself — is far more reliable than any app verdict.
What should I do if someone made a fake video of me?
Save evidence, report it to the platform, and report it to the authorities. Do not engage with the account spreading it, and ask the platform to remove it. If the fake is used for blackmail, do not pay — paying almost never ends the demands. Our data breach recovery guide covers the same evidence-first approach for other online attacks.
Are video calls safe now?
Video calls with people you know, on accounts you already trust, are still safe — the risk comes from unexpected calls and messages that demand urgent action. The rule is simple: a familiar face does not replace verification when money, codes or secrecy are involved.
Will deepfakes keep getting harder to spot?
Yes, and that is exactly why the habit matters more than the clues. Visual signs will fade as the technology improves, but the verification rule does not age: pause, contact the person yourself through a channel you trust, and never send money or codes on the strength of a single call or clip. Build that habit now, teach it to your family, and deepfakes lose the one thing they depend on — your rushed trust.