How to Spot a Deepfake Video: Signs and Checks

A clip of a company founder saying something shocking starts circulating. A colleague forwards a video of a public figure endorsing an investment. A video call with "the CFO" includes an unusual payment request. Before you share, reply or pay, it is worth knowing how to spot a deepfake video. This guide walks through the context checks, visual signs and audio clues to look for, a simple step-by-step routine, and the tools that can help when your eyes are not enough.

Why spotting a deepfake video is harder than it looks

Deepfake tools have improved quickly. Singapore's Cyber Security Agency (CSA) warns that AI is producing deepfakes that can be indistinguishable even to the trained eye, and researchers behind the MIT Media Lab's Detect Fakes project note that there is no single tell-tale sign of a fake. The FBI adds that generative AI can correct for the human errors that might otherwise serve as warning signs of fraud.

That does not mean you are powerless. Most deepfake videos still give something away, either in the frames themselves, in the audio, or in the story around them. The trick is to check all three rather than relying on one.

Step one: check the context before the pixels

The CSA's "3A" approach begins with assessing the message: its source, its context and its aim. These questions often expose a fake faster than any frame-by-frame analysis.

• Source. Who first posted the video? Is it on the person's or organization's official channels, or on an account you cannot trace?
• Context. Does the content fit what you know? A leader announcing major news only in an unverified clip, or a public official promoting a private investment scheme, should raise doubts.
• Aim. Is the video pushing you to do something urgent, unsafe or unusual, such as clicking a link, sharing personal information, sending money or buying something that sounds too good to be true?

Fact-checkers at PolitiFact recommend lateral searching: opening several tabs to see what other sources say about the claim and who is sharing it. If a dramatic statement from a well-known person appears nowhere else, treat it with suspicion.

Visual signs of a deepfake video

If the context does not settle it, look closely. Watch at full screen, pause often and step through the clip slowly. These are the areas public guidance most often points to.

1. Edges of the face

Look for blurring around the edges of the face, the hairline or the side profile, and for uneven resolution between the face and the rest of the frame. The CSA lists unnatural edges around facial features as a common sign.

2. Skin texture and age

The MIT Media Lab suggests checking whether the cheeks and forehead look too smooth or too wrinkly, and whether the apparent age of the skin matches the hair and eyes. An "airbrushed" face on an otherwise sharp video is worth a second look.

3. Eyes, eyebrows and blinking

Does the person blink too little or too much? Are shadows around the eyes and eyebrows where you would expect them? Inconsistent light reflections in the eyes are another sign the CSA highlights.

4. Lighting, shadows and glasses

Deepfakes may fail to reproduce the physics of a scene. Check whether shadows on the face match the light in the room, and whether glare on glasses is missing, excessive, or fails to change angle when the head moves.

5. Facial hair and moles

Beards, moustaches, sideburns and moles can look painted on, flicker, or appear and disappear between frames.

6. Teeth, hands and accessories

The FBI's list of imperfections to look for includes distorted hands, unrealistic teeth or eyes, indistinct or irregular faces and unrealistic accessories such as glasses or jewelry.

7. Background and movement

Look for blurred, out-of-focus or warped areas in the background, and for movements that look unnatural, especially when the head turns or something passes in front of the face.

Audio signs to listen for

Many deepfake videos pair a manipulated face with a cloned or re-timed voice. Use headphones and listen for:

• Lip sync. Lips that drift out of sync with speech are one of the most widely cited signs, and some deepfakes are built specifically on lip-syncing.
• Flat delivery. The CSA mentions limited variance in tone and inflection.
• Background noise that does not fit. The CSA lists incongruent background noise, such as room sound that does not match the setting, as a warning sign.
• Tone and word choice. The FBI advises listening closely to tone and word choice to tell a genuine call from a cloned voice.

A step-by-step routine for checking a suspicious video

1. Pause before you react. Do not share, reply or act on the clip yet.
2. Trace the original. Find the first upload and check the account's history. Search for the claim on trusted news sites and the person's official channels, and run a reverse-image search on a screenshot of a key frame to find earlier copies, as PolitiFact suggests.
3. Watch it slowly. Step through key moments frame by frame, focusing on the face, eyes, mouth and edges.
4. Listen separately. Replay the audio with headphones and watch the lips against the words.
5. Look for provenance labels. Some content carries Content Credentials, an open standard from the Coalition for Content Provenance and Authenticity (C2PA) that records a file's origin and edits. The CSA notes that platforms have also announced metadata tags or watermarks to indicate AI-generated content.
6. Run a detection check. Upload a short clip to DeepWall's free deepfake video detection tool. It accepts MP4, MOV or WEBM files of up to 30 seconds and 50 MB, is built to look for AI-generated frames and face swaps, and needs no sign-up. For a still frame or screenshot, use the free deepfake image detection checker. If the concern is the voice, you can upload an audio file on the deepfake audio detection page.
7. Confirm with the person. Contact the person or organization shown, using contact details you already have rather than any provided in the video or message.

Treat a detection score as one signal, not a verdict. Scores are probabilities from AI models, and heavy compression, low resolution, small or turned-away faces and screen recordings all reduce accuracy.

Deepfakes on live video calls

Deepfakes are not limited to clips posted online. The FBI warns that criminals generate video for real-time chats with alleged company executives, law enforcement or other authority figures. The CSA describes a February 2024 case in which a finance worker at a multinational firm paid US$25 million to scammers who used deepfake technology to pose as the company's chief financial officer on a video conference call.

On a live call you cannot pause and replay, so process matters more than perception:

• Verify any payment or credential request through a second, known channel, such as calling back on a number you already have. The FBI recommends hanging up and calling the organization directly.
• Ask a question only the real person would know the answer to, as the CSA suggests.
• Agree on a verification word or phrase in advance for high-risk requests, a step the FBI recommends to families.
• Make it normal for staff to pause and verify an urgent request from a senior leader without fear of getting it wrong.

When the fake shows you or your leadership team

Spotting a deepfake someone sends you is one problem. Finding out that a fake video of you or your executives is circulating somewhere you are not looking is another. Manual searching does not scale, and by the time a clip reaches you it may already have spread.

DeepWall's deepfake detection software is built for that second job. You enroll the people you want to protect with 3 to 5 photos and a 30-second voice sample. DeepWall then checks newly uploaded videos on YouTube, X, TikTok, Facebook, Instagram, Reddit and news sites with embedded video, matching faces frame by frame and checking the audio for cloned voices. When a video crosses your match threshold, you get an alert, typically within minutes of detection. Professional and Executive plans add evidence export and takedown templates for reporting the video to the platform.

Plans start at $49 per month, and every plan includes a 30-day free trial with no credit card required. Start your free trial, or learn more about what a deepfake is and how deepfakes are made.

Frequently asked questions

What is the easiest way to spot a deepfake video?

Start with the context: check who posted it, whether trusted sources report the same thing and whether it pushes you to act urgently. Then look for visual and audio signs such as blurred face edges, unnatural blinking, mismatched lighting and poor lip sync.

Can you always tell a deepfake by looking at it?

No. Public guidance warns that high-quality deepfakes can be very hard to tell apart from real footage, even for trained viewers. Combine close viewing with source checks, detection tools and direct confirmation from the person shown.

Is there a free tool to check if a video is a deepfake?

Yes. DeepWall's free deepfake video detection tool lets you upload a clip of up to 30 seconds and 50 MB in MP4, MOV or WEBM format with no sign-up. Treat any result as one signal alongside other checks.

How do I know if a video call is a deepfake?

Be cautious with any unexpected request for money, credentials or sensitive data. End the call and verify through a channel you already trust, such as a known phone number, or ask a question only the real person could answer.

What should I do if I find a deepfake video of me?

Do not share it. Save the URL, the posting account, the date and a copy of the video, report it to the platform, and warn colleagues who might be targeted. If money was lost to fraud in the United States, the FBI asks victims to report it to its Internet Crime Complaint Center (IC3).