Deepfake Awareness
How to spot AI-generated fakes — and the verification habits that protect you
These are the most reliable visual/audio cues — but technology improves fast. Verification habits (next tab) matter more than detection.
Unnatural blinking or eyes
Early deepfakes blinked rarely or at wrong intervals. Better models are improving, but eye movement and eyelid edges still look slightly wrong on close inspection.
Lips out of sync with speech
Audio often doesn't perfectly match lip movement, especially at the start and end of words. This is clearest on video calls or screen recordings.
Hair and ear edges
AI struggles with fine details around the hairline and ear outline. These edges often look blurred, smeared, or flickering.
Inconsistent lighting
The face may have slightly different lighting than the background or body. Shadows don't always match the environment.
Voice sounds slightly 'smoothed'
AI voice clones often sound slightly too clean or have unnatural pauses. Background noise or room acoustics may not match.
Extreme expressions look distorted
When the person laughs, cries, or makes a strong expression, the distortion is often most visible at the face edges and neck.
Tool tip
For suspicious videos: try the InVID/WeVerify browser extension or search the thumbnail on Google Images to find the original source.