Straight answers on how deepfakes work, how to spot them, and what to do if you find one.
A deepfake is synthetic media where a person's face, voice, or body is replaced or generated by artificial intelligence. The technology works by training a neural network on hundreds or thousands of images and audio clips of a target person.
The AI learns the person's facial expressions, lip movements, and vocal patterns, then generates new content that never actually happened. Most deepfakes use a technique called generative adversarial networks (GANs), where two AI models compete against each other until the output is convincing.
Anyone with a decent GPU and free software can produce a basic deepfake in under an hour. The barrier to entry drops every year. Read more about how detection technology is keeping up.
Start with the eyes. Deepfake subjects often blink at unnatural intervals or not at all. Watch the lip sync closely, especially on hard consonants like B, P, and M. If the lips don't fully close on those sounds, something is off.
Check the lighting on the face against the background. Deepfakes frequently get the direction and colour temperature of light sources wrong. Look at the edges of the face, particularly around the jawline and hairline, where the AI-generated face meets the original footage.
Play the video at half speed. Artefacts that are invisible at normal speed become obvious when slowed down. For a full breakdown, see our guide to spotting deepfakes.
Look for these specific tells: warped or asymmetric ears, teeth that blur or shift between frames, skin texture that looks too smooth or waxy, jewellery or glasses that flicker or distort when the subject moves, and hair that appears painted on rather than individual strands.
Backgrounds near the subject's head often warp or shimmer. Skin tone may shift slightly when the head turns. Earrings, collars, and necklines frequently glitch because the AI struggles with accessories at the boundary of the generated face.
None of these are guaranteed proof on their own, but three or more red flags in one video should raise serious suspicion. Try the free awareness scanner to test your detection skills.
Yes, several free options exist. Microsoft Video Authenticator analyses photos and video frames, giving a confidence score for manipulation. Sensity AI offers a free tier for detecting AI-generated faces. Hive Moderation provides an API and browser tool for detecting AI-generated images. AI or Not is a simple drag-and-drop checker for images.
For reverse-image searching suspicious photos, use Google Images, TinEye, or Yandex Images.
No single tool catches everything. The best approach is to combine a detection tool with manual checks for visual red flags. Run the content through two different tools before drawing a conclusion.
Deepfakes weaponise trust. When you can no longer believe what you see and hear, every piece of media becomes suspect.
Scammers use deepfake video calls to impersonate executives and authorise wire transfers. In 2024, a Hong Kong finance worker transferred $25 million after a deepfake video call with what appeared to be their CFO. Non-consensual intimate imagery destroys lives and is used for blackmail. Political deepfakes can shift election outcomes by going viral before fact-checkers respond.
Voice cloning lets criminals impersonate family members in emergency scam calls. The damage is real, immediate, and often irreversible. Report deepfake abuse to the eSafety Commissioner in Australia.
Reduce the raw material available to attackers. Set social media profiles to private and avoid posting high-resolution face photos publicly. The more images of your face that exist online, the easier it is to train a model on you.
Establish a family code word for emergency phone calls so you can verify identity if someone claims to be a relative in distress. Enable two-factor authentication on all accounts so that even a convincing voice clone cannot bypass your security. Watermark professional photos and videos you publish.
If you discover a deepfake of yourself, report it to the eSafety Commissioner (esafety.gov.au), the platform hosting it, and your local police. Document everything with screenshots before the content is removed. Read our full protection guide for more steps.