How to Spot a Deepfake Video: A Family Guide

How to spot a deepfake video: the short answer

Learning how to spot a deepfake video now means checking the source, not just staring at faces. The FBI says AI-generated content is often hard to identify. So pause before reacting, verify the person through a channel you already trust, and look for AI labels or Content Credentials. Finally, use a family code word for urgent requests.

Knowing how to spot a deepfake video used to mean looking for odd blinking or warped hands. However, that advice is fading fast. In December 2025, the FBI warned that AI-generated content “has advanced to the point that it is often difficult to identify.”

That does not mean families are helpless. Instead, the safest way to spot a deepfake video has simply changed. Official guidance from the FBI, the FTC and regulators in the UK and Australia now puts verification first and visual clues second.

This guide therefore pulls that guidance together in plain English. It covers the visual clues that still help, the labels and provenance tools you can check, and a simple routine for the whole household. It also explains where to report a fake. Note that we have not tested detection tools for this article. Everything here comes from the public sources listed at the end.

What is a deepfake video?

A deepfake video is footage that uses AI to make a real person appear to say or do something they never did. Sometimes AI creates the whole clip. In other cases, a real video gets a swapped face or a cloned voice.

Admittedly, many deepfakes are jokes, satire or harmless experiments. However, criminals also use them. The FBI’s December 2024 public service announcement described criminals using AI-generated video, images and voices for fraud. Common examples include fake emergencies, fake celebrity investment pitches and impersonated officials.

Can you still spot a deepfake video by looking closely?

Sometimes, but you should not rely on your eyes alone to spot a deepfake video. In 2024, the FBI listed visual clues such as distorted hands or feet, unrealistic teeth or eyes, and indistinct faces. Australia’s eSafety Commissioner adds blurring around the face, odd skin texture, lighting glitches and poorly synced audio.

Yet experts warn these tells are disappearing. Siwei Lyu, a deepfake researcher at the University at Buffalo, wrote in January 2026 that modern models produce stable faces without the old flicker or warping. He also said that “simply looking harder at pixels will no longer be adequate.”

Research also points the same way. A January 2026 study in Cognitive Research: Principles and Implications found that people judged still images at about chance level. Interestingly, people did better with videos than the detection algorithm did. Even so, nobody should treat a gut feeling as proof.

Visual and audio clues worth noticing

In short, treat these as reasons to slow down, not as a verdict:

Blurring or “seams” where the face meets hair, ears or neck.
Lighting on the face that doesn’t match the room.
Lips slightly out of sync with the words.
Lag, freezing or strange movement on a live video call.
A voice that sounds right but has an odd rhythm or flat emotion.
Teeth, eyes, hands or jewellery that look unnatural.

Remember the flip side, too. For example, a clean, flawless clip proves nothing. Likewise, a real video can look odd because of poor lighting or compression.

Why is checking the source the best way to spot a deepfake video?

Every official source we read makes the same point in different words. So the most reliable answer to how to spot a deepfake video is to check where it came from and whether the story holds up.

The FTC puts it simply in its alert on family emergency scams: “Don’t trust the voice.” Instead, call the person on a number you know, or check with another relative. The FBI gives similar advice. It says to hang up, look up the organisation’s number yourself, and call it directly.

Above all, this approach works even when the fake is perfect. After all, a scammer can copy a face, but they cannot easily answer a call to your sister’s real phone number.

Can AI labels help you spot a deepfake video?

Several platforms and AI companies now label or tag AI-generated content. These signals can help you spot a deepfake video when they are present. Even so, missing labels prove nothing.

Signal	What the company says it does	Limits
YouTube label	Creators must disclose realistic altered or synthetic content; YouTube may also add labels itself	Depends largely on honest disclosure
Meta “AI info”	Labels content using industry signals, self-disclosure and Meta’s own tools	Meta has admitted mislabelling minor edits in the past
TikTok AI label	Requires labels on realistic AI content and reads Content Credentials	TikTok says labels can disappear when people re-upload content
OpenAI Sora	Visible watermark plus embedded C2PA data on outputs	People can crop or edit out watermarks
Google SynthID	Invisible watermark on content from Google’s AI tools, with checks in the Gemini app	Only detects Google’s own watermark

How Content Credentials work

Content Credentials are based on an open standard from the C2PA coalition. They record where a file came from and how someone edited it. On sites that support them, a small “CR” icon then shows that record. You can also upload a file to the free Verify tool linked from contentcredentials.org.

There is a catch, though. For instance, the C2PA FAQ admits that credentials embedded in a file “can be separated” from it. Similarly, the UK regulator Ofcom found in 2025 that basic edits can often strip watermarks. As a result, a video with no credentials is not automatically fake, and a real-looking video is not automatically real.

Checking with Google tools

First, Google added SynthID checks to the Gemini app for images, video and audio. In May 2026, Search Engine Journal reported that Google was bringing these checks to Search through Lens and Circle to Search. However, coverage differed on which media types Search supports, so check Google’s current help pages. Above all, remember that SynthID only finds Google’s own watermark.

A five-step routine to spot a deepfake video at home

Families do not need special software to spot a deepfake video. Instead, agree on a short routine everyone can follow, including children and older relatives.

Pause. Urgency is the scammer’s main tool. If a video demands money, secrecy or speed, stop before you act.
Verify through another channel. Call the person on a saved number, or contact the company through its official website.
Use a family code word. The FBI recommends a secret word or phrase to confirm identity. Our guide to family code words for AI voice scams explains how to set one up.
Check the source. Find the original upload. Then look at the account’s history, and check whether trusted news outlets report the same thing.
Look for labels and credentials. Check platform AI labels, the “CR” icon or a provenance tool. Still, treat their absence as neutral.

You can also save this short version to the family group chat:

If a video or call asks for money, codes or secrecy: stop, hang up, call back on a number we already have, and ask for our code word. Never send money, gift cards or crypto because of a video alone.

Which deepfake scams should families know about?

The FBI and FTC warnings also describe several recurring patterns. As a result, knowing them makes it easier to spot a deepfake video in context.

Family emergencies. A cloned voice or face claims to be a relative in trouble and asks for urgent money.
Impersonated officials. The FBI warned in May 2025 about AI-generated voice messages that pretend to come from senior US officials. It later said the campaign also targets officials’ families.
Fake investment pitches. A famous face appears to endorse a crypto or trading scheme.
Live video call fraud. A colleague or boss appears on a call and asks for a payment.
Fake apps and links. A deepfake ad points to a download or login page.

For more on the last two, read our guides to AI agent scams and how to spot a fake AI app.

How can you protect your own face and voice?

Of course, you cannot stop every misuse. Still, the FBI suggests limiting how much of your image and voice is public. In practice, that means:

Making social media accounts private where possible.
Reviewing who can see old videos, especially of children.
Thinking twice before posting long clips of your voice.
Never sharing one-time passcodes, even with someone who looks and sounds familiar.

Where can you report a deepfake?

Reporting helps platforms and police spot patterns. Where you report, however, depends on what happened:

Situation	Where to report (as of October 2026)
Fraud or scam in the US	FBI at ic3.gov and the FTC at ReportFraud.ftc.gov
Fraud in England, Wales or Northern Ireland	Report Fraud, which replaced Action Fraud (in Scotland, call Police Scotland on 101)
A fake on a social platform	The platform’s own in-app reporting tools
Explicit images of someone under 18	NCMEC’s Take It Down service and the CyberTipline
A US platform ignoring a removal request for intimate images	The FTC, under the TAKE IT DOWN Act

On that last point, the FTC says the TAKE IT DOWN Act now requires covered platforms to remove reported non-consensual intimate images, including AI-made ones, within 48 hours. Those platform duties took effect on 19 May 2026. Meanwhile, the UK has created offences covering intimate deepfakes. If you or a family member are a target, contact the police as well as the platform. Our guide to whether AI agents are safe covers related privacy habits.

Who most needs to know how to spot a deepfake video?

Everyone benefits from knowing how to spot a deepfake video. Even so, some groups face more risk:

Older relatives, whom scammers may target with fake grandchild emergencies.
Teenagers, who face image-based abuse and fake “friend” accounts.
Small-business owners and finance staff, who may receive fake payment requests on video calls.
Anyone with a public profile, whose face and voice are easy to copy.

Finally, keep the conversation calm. The goal is a household habit of checking, not fear of every video online.

Key takeaways

The FBI says AI-generated content is now often difficult to identify by eye.
Verifying through a trusted second channel beats hunting for visual glitches.
AI labels, watermarks and Content Credentials help when present, but edits can strip them.
A family code word and a “pause, call back” rule stop most emergency scams.
Report scams to the FBI, FTC or your national fraud service, and report fakes to the platform.

How to spot a deepfake video: FAQs

What are the signs of a deepfake video?

Watch for blurring around the face, mismatched lighting, lips out of sync and lag on live calls. However, the FBI says AI fakes are often hard to identify, so treat signs as a reason to verify.

How can I tell if a video call is a deepfake?

Watch for lag, a voice that doesn’t match the face and odd movement. Then hang up, call back on a number you already have, and ask for your family code word before sending any money.

What is the quickest way to check if a video is AI-generated?

Check for a platform AI label and for Content Credentials, or use the Gemini app’s SynthID check. Still, a missing label proves nothing, because edits can strip watermarks.

Can AI tools detect deepfakes better than people?

Not reliably. A 2026 study found algorithms excelled on still images, but people did better on videos. Also, SynthID only finds Google’s own watermark.

Are deepfake videos illegal?

Not automatically, but using one for fraud is a crime. The US TAKE IT DOWN Act covers intimate deepfakes, and the EU requires deepfake disclosure from August 2026. This is not legal advice.

What should I do if someone made a deepfake of me?

Save links and screenshots, report it to the platform and contact the police. For anyone under 18, use NCMEC’s Take It Down service and the CyberTipline.

Can someone make a deepfake of me from social media?

Yes, the FBI warns that criminals use public photos and voice clips. So make accounts private and limit long video or voice clips.

How to Spot a Deepfake Video: A Practical Guide for Families

What is a deepfake video?