What a deepfake actually is
A deepfake uses AI to swap a face, copy a voice or generate a whole scene. Some are harmless jokes or clearly labelled creative edits. Others are made to deceive, such as a celebrity 'endorsing' an investment, a politician 'saying' something shocking or a fake video of a private person.
The technology improves quickly, so the goal is not to become a perfect detector. It is to build the habit of checking before trusting.
Visual clues that still help
Watch the mouth: words may not match lip movements exactly. Look at edges where the face meets hair, ears or glasses, which may blur or flicker. Hands, teeth, jewellery and background text can look warped. Lighting on the face may not match the rest of the scene.
Slow the video down or pause on several frames. Problems that are invisible at full speed often become obvious when you look closely.
Listen as well as look
AI voices can sound flat, oddly paced or slightly robotic, with strange breathing or no background noise at all. Emotional moments may sound too even. A familiar voice that is not quite right is worth a second check.
Voice clones are also used in scams outside social media, such as calls claiming to be a relative in trouble. If a voice asks for money urgently, hang up and call the person back on a known number.
Context is your best tool
Ask who posted the clip first, when and why. Is it on the person's official account or reported by reliable news outlets? Does the account sharing it have a history of sensational content? Search for the claim itself, not just the video.
Many platforms label some AI-generated content, and creators are often asked to disclose realistic AI edits. A label helps, but missing labels do not prove a video is real.
What to do with a suspected deepfake
Do not share it, even to say 'is this real?', because that spreads it further. Report it if it impersonates someone, promotes a scam or harms a private person. If you are the target of a deepfake, save evidence, report it on the platform and seek support from people you trust.
Frequently asked questions
Are all AI videos deepfakes?
No, many AI videos are open creative projects, filters or clearly labelled edits. The concern is content designed to make people believe something false or to harm someone.
Can detection apps tell me for sure?
Detection tools can help, but none are fully reliable and results can be wrong. Combine them with context checks and trusted sources rather than relying on one tool.