How to Spot a Deepfake Video: 7 Signs Anyone Can Check

Safety & scams Tutorial7 min read·Updated July 4, 2026
The short answer

Deepfake videos use AI to swap or animate a person's face, but they leave telltale signs. Look at the edges of the face, whether the eyes blink naturally, and whether the mouth matches the audio. No single sign is definitive, but two or three together strongly suggest the video has been manipulated.

Deepfake videos have gotten more convincing, but they still leave clues. You do not need any special tools or technical background to catch most of them — you just need to know where to look. These seven signs can each be checked in seconds, on any device, with no downloads required.

Look at the face edges

Watch the border between the person's face and their hair, neck, or background. In deepfakes, this edge often looks slightly blurry, smeared, or wobbly — especially when the person moves their head. Real faces have a clean, consistent boundary with the surrounding scene.

If the skin seems to float, shimmer, or flicker at the edges during movement, that is a strong warning sign. Pause the video on a frame where the person is mid-motion to make this easier to see.

Watch how the eyes blink

Natural blinking is irregular — sometimes fast, sometimes slow, occasionally skipped for a moment while someone is focused. Early deepfakes blinked too rarely or not at all; newer ones blink more, but the blinking still sometimes feels mechanical or too uniform.

Watch for eyes that blink at an identical rate throughout the clip, or eyes that look slightly glassy or flat compared to the rest of the face. Natural eyes also catch light and move slightly together — both things AI sometimes gets wrong.

Check if the mouth matches the audio

Play the video with sound and watch the lips closely. In a real video, the exact shape of the mouth matches each sound being spoken. In deepfakes, the lip movements sometimes run slightly ahead or behind the audio, or the shapes do not quite match the words being said.

This is easiest to catch on hard consonants — the letters "p," "b," and "m" require the lips to press together fully. If the lips are open when you hear one of those sounds, the sync is off.

Look at the lighting on the face versus the background

Lighting in a real video is consistent — if the room is lit from the left, the face should have a shadow on the right side, and everything in the scene should follow the same logic. In deepfakes, the lighting on the swapped face often does not quite match the rest of the scene.

The face might look a bit too evenly lit, or it might cast a shadow in a direction that does not match the background. This mismatch is one of the harder things for AI to fake convincingly.

Check hair, glasses, and jewelry

Small accessories and fine details are genuinely difficult for AI to maintain consistently throughout a video. Watch for earrings that blur into the neck or flicker, glasses frames that warp where they cross the eye, necklaces that seem to meld into clothing, or individual strands of hair that look oddly smooth or plastic.

These elements should remain stable and look natural throughout the whole clip. Any flickering or inconsistency in a small detail is worth noting.

Watch the background near the head

When the person moves their head, watch the area of the background immediately around and behind them. Deepfakes sometimes cause the background to warp, ripple, or blur in a band that follows the edge of the head during movement.

This is most visible during quick turns or nods. Pausing on a frame just after a fast head movement often makes this distortion easy to spot.

Ask where the video came from

No visual check replaces source checking. Before accepting a shocking or surprising video as real, ask: who posted this, on what account, and why? A video of a celebrity making an unexpected statement that appears on a new or unfamiliar account, shared without context, deserves extra scrutiny.

Real breaking news and genuine viral moments are picked up by many sources quickly. If only one account is sharing a clip and it has no clear origin, search for the same content elsewhere before treating it as real.

When Several Signs Point the Same Way

No single sign is proof on its own — some appear in low-quality real videos too. But when you notice two or three of these signs in the same clip, the probability of manipulation goes up sharply.

The goal is not to be completely certain before you act — it is to build the habit of pausing to look. That moment of attention is the most useful thing you can practice.

What to try next: Fake videos are one part of the picture. See How to Spot AI-Generated Photos for the same approach applied to still images, and read AI Voice Cloning Scams to understand the voice-based version of these threats.

Published July 4, 2026 · Updated July 4, 2026How we test →

Frequently asked questions

What is a deepfake video?
A deepfake video uses AI to place one person's face onto another person's body, or to animate a still photo so it appears to speak. The result can look surprisingly realistic, especially in short clips.
Do I need special software to detect deepfakes?
No. These seven signs require only your eyes and a willingness to pause and look carefully. Some online detection tools exist but are not always reliable, and the signs here work without any tools at all.
Are all deepfakes easy to catch?
No. Quality varies widely. Some are poorly made and obvious; others are very convincing. Checking context — where the video came from and why it is being shared — matters as much as visual signs.
Can deepfakes be used in scams?
Yes. Scammers use deepfake videos to impersonate executives, celebrities, or even family members to trick people into sending money or sharing personal information. Video calls are also being faked this way.
Should I report a deepfake video?
Yes. If you believe a video is fake and is being used to deceive people, report it to the platform where you found it. If it involves fraud or a crime, report it to local law enforcement as well.
Radim Sekera
Founder & editor

Radim is a software developer who spends his days building with AI and his evenings explaining it to family members who don’t care how it works — only what it can do for them. Every guide is tested by hand before it’s published.