Deepfake Video Scams — How to Spot Fake Faces

In 2024, a finance worker at a Hong Kong multinational wired US$25 million after a video call that appeared to feature his chief financial officer and several colleagues. Every face on the call was a deepfake. Every voice too. The entire meeting had been synthesised by AI using public video footage of the real employees.

That case was a corporate fraud. But the same technology now runs on consumer-grade software, and families have started seeing it too. Seniors have received WhatsApp video calls that appear to be from a grandchild, asking for emergency money. The face looked right. The voice sounded right. It was not real.

This guide covers what deepfake video fraud looks like at home, the five visual tells you can learn to spot in ten minutes, and the single safeguard that beats every detection trick: a family code word.

How this scam is different from voice cloning

Our earlier guide on voice-clone scams covered audio-only fakes. Video deepfakes are the next step. They use the same underlying AI but add synthetic video of a person's face matched to synthesised speech.

The scammer usually gets 30 seconds of source material — a Facebook video, an Instagram reel, a WhatsApp status — and generates a new video with that person saying whatever the scammer types. The output is not perfect yet. That imperfection is what you will learn to spot.

The five visual tells

You do not need special software. You need to know where to look.

1. Watch the eyes

Real eyes blink at irregular intervals — typically 15 to 20 blinks a minute, but not evenly. AI-generated faces tend to blink at unnatural rhythms: either too rarely, or in a metronome-like even pattern. Watch for 30 seconds. If the blink pattern feels mechanical, that is a signal.

Also look at eye reflections. Real eyes reflect the light source in the room — a window, a lamp, a computer screen. Deepfake eyes often have missing or mismatched reflections. If one eye reflects a bright light and the other does not, something is wrong.

2. Check the lip-sync

When real people speak, the mouth movement and the sound are perfectly matched. Deepfakes often have a tiny delay — maybe 100 milliseconds — between lip movement and audio. Your eye catches it as "something is off" before you can name it.

Ask the person on the call to say a word with sharp consonants, like "Popcorn" or "Pineapple." Watch the lips. In real video, the lips close firmly on the P. In a deepfake, the closure is often mushy or slightly out of sync.

3. Look at the hair and the edges of the face

Deepfakes struggle with fine detail around the hairline, the ears and the jaw. Look for:

Hair that moves strangely or appears to "float" against the background
A slight blur or ghosting at the edge of the face when the head turns
Ear shapes that look unusually symmetrical — real ears are rarely identical

4. Check the skin texture

Real skin has pores, small lines, micro-shadows. AI-generated faces often have skin that is too smooth — like a photo retouched for a magazine. Ask the person to come closer to the camera. A deepfake will show an unnatural smoothing when zoomed in.

5. Ask them to turn their head and look away

Many deepfakes work best when the face is pointed straight at the camera. Ask them to turn their head 90 degrees, or look over their shoulder at something behind them. If the face distorts, breaks up, or the edges ghost, that is the most reliable tell of all.

You can also ask them to wave a hand slowly in front of their face. Deepfakes usually do not handle occlusion well — you will often see the hand flicker, or the face will briefly distort when the hand crosses it.

The one safeguard that beats all of this — a family code word

Visual tells help, but they are not foolproof. Deepfake quality improves every few months. The method that works reliably — and will keep working — is the old-fashioned one: a shared secret.

Agree on a code word with each of your closest family members. One word per person. Something unguessable from social media — not a pet's name, not a hometown, not a birthday. Something that would never appear anywhere online.

The rule: Any urgent request for money, however genuine-looking, must be confirmed with the code word.

If your grandson calls sobbing that he has been in an accident and needs ₹50,000 now, and he cannot say the code word, the call is a scam — no matter how real the face and voice look.

Teach this to every family member. Write the code word down. Tell everyone that no one will ever be offended if you ask for the code word before helping.

Tools consumers can use

A few detection tools exist, though none is perfect:

Intel FakeCatcher — uses blood-flow analysis of facial pixels to detect synthesis. Available in limited products.
Reality Defender — primarily a business tool; a consumer version was announced in 2025.
Microsoft Video Authenticator — older tool, not actively maintained.

For most people, the right approach is not to rely on tools. The code word beats every tool. Tools catch 80–95% of deepfakes; a code word catches 100%.

If you are already on a suspicious call

Stay calm. Do not hang up immediately — scammers will try again and may escalate.

Ask the code word. If they cannot say it, the call is a scam.
Ask them to wave a hand in front of their face or turn sideways. Watch for distortion.
Say you will call back on their known number — not the number they are calling from. Hang up.
Call the real person directly on a number saved in your phone. If they answer normally, that confirms the first call was a scam.
Report to cybercrime.gov.in (India), reportfraud.ftc.gov (US), or Action Fraud (UK).

Teach your family this week

Sit with your spouse or an adult child this Sunday. Agree on a code word. Test it in a practice call. Write it down somewhere private.

Teach your grandchildren too. If your grandson knows that Grandma will always ask for the code word before sending money, he will also know to never be surprised by it in a real emergency.

Frequently Asked Questions

How common are deepfake video scams in 2026?

They are still less common than phone-based voice scams, but growing rapidly. Reported cases are mostly in higher-value frauds — business email compromise, high-net-worth individual fraud — but consumer-targeted cases have started appearing in 2025 and 2026.

Is there an app that can detect deepfakes for me?

A few exist (Intel FakeCatcher, Reality Defender) but none is reliable enough for consumer use without training. For home use, the family code word is far more effective than any detection tool.

What is a family code word?

A secret word agreed between close family members, used to verify identity in urgent or suspicious calls. It should be something that would never appear on social media — not a pet's name, hometown or birthday. When someone calls asking for money urgently, you ask for the code word before helping.

Can deepfakes impersonate voices as well as faces?

Yes. Modern deepfakes combine synthetic voice (from voice cloning) with synthetic face (from video generation). The scammer types what they want the person to say, and the software produces matched audio and video. This is why the code-word defence is more reliable than any visual or audio check.

What should I do if I think I have spoken to a deepfake?

Hang up. Call the real person back on their known number immediately to confirm. If you sent money, call your bank's fraud helpline within the hour, and file a complaint at cybercrime.gov.in (India) or reportfraud.ftc.gov (US).

Are video calls on WhatsApp safer than on unknown apps?

WhatsApp end-to-end encryption protects the content from being intercepted, but it does not protect against a scammer's identity being faked. A deepfake WhatsApp video call is technically possible. The app does not change the risk — the code word remains the best defence.

Keep reading

✅ Reviewed & Verified by Eleanor Shaw | techfor60s.com Editorial Desk

Last fact-checked: 2026-04-18

Next scheduled refresh: 2026-07-18