How to Spot a Deepfake: 8 Signs and the Tools That Help
Deepfakes - face-swapped and AI-synthesized images and video - are now good enough to fool a quick glance. But most still leave subtle clues if you know where to look. Here are eight signs to check, plus a fast way to confirm your suspicion with a detector.
The quick answer
To spot a deepfake, slow the media down and study the face boundary, eyes, lighting and audio sync - the areas manipulation tools handle worst. Then run the image through a deepfake detector for a probability score. Use the two together, since neither is conclusive alone.
Check a face now
Upload a photo and our free deepfake detector scores it for face-swap and synthesis artifacts in seconds.
Open the Deepfake Detector →8 signs of a deepfake
1. The edges of the face
The boundary where the face meets hair, ears and neck is the weakest point. Look for blurring, flickering or a faint seam - most visible when the person turns their head.
2. Eyes and blinking
Watch for unnatural or infrequent blinking, eyes that do not quite track together, and reflections in the eyes that do not match the scene's light sources.
3. Skin tone and texture
Swapped faces often have a slightly different colour or smoothness than the neck and ears around them. Look for a mismatch at the jawline.
4. Lighting and shadows
Check that the light on the face matches the rest of the scene. Deepfakes frequently get shadow direction and intensity subtly wrong.
5. Teeth, hair and fine detail
Individual teeth can blur into a single block, and stray hairs often smear or shimmer between frames in video.
6. Mouth and audio sync
In video, watch the lips against the words. Deepfakes often have small timing mismatches, and mouth shapes that do not fully match the sounds.
7. Head and neck movement
An unnaturally still head on a moving body, or a face that seems to float slightly against the neck, is a red flag.
8. Context and source
Who posted it, and where did it first appear? A shocking clip from an unverified account with no original source is worth doubting before you even look closely.
How to check with a deepfake detector
A deepfake detector analyzes a face for the artifacts that face-swap and synthesis techniques leave behind, and returns a probability - not a yes/no.
- Take a clear frame. For video, screenshot a sharp, front-facing frame of the face.
- Upload it to the detector.
- Read the score. A high percentage suggests manipulation; a low one suggests an authentic face. Middle scores are inconclusive - check another frame.
Detectors work best on clear, front-facing faces. Tiny, low-resolution or heavily compressed faces are much harder to assess.
The limits of deepfake detection
- False positives: genuine but heavily filtered or edited photos can be flagged.
- Novel methods slip through: a brand-new technique may leave traces the detector has not learned.
- Compression hurts accuracy: social platforms re-encode media, weakening the signal.
For anything consequential - a suspicious video of a public figure, a dating-profile photo, an identity document - treat the detector as one input, corroborate the source, and do not rely on a single check.
Suspect a deepfake?
Run the face through our free deepfake detector - no signup, instant score.
Try the Deepfake Detector →Frequently asked questions
How can you spot a deepfake?
Look for unnatural blinking, blurry or flickering face edges, mismatched skin tone, lighting that does not match the scene, and audio that is out of sync with the lips - then confirm with a detector.
Are deepfake detectors accurate?
They are a strong signal, not proof. They work well on many face swaps but can be fooled by high-quality manipulations and can flag heavily edited real footage. Combine with visual checks and context.
What is the easiest sign of a deepfake?
The face boundary. Watch the edges of the face, hairline and neck for blurring or a faint seam, especially when the head turns.
Related: How to Tell if an Image Is AI-Generated · Deepfake Detector · AI Image Detector