How to detect a deepfake
Updated: August 2026 · By César Cerrudo
For decades, a photo or a video carried enormous power: "If it's on camera, it happened."
Artificial intelligence is changing that relationship. A video can apparently show a person saying something they never said. An image can depict an event that never happened. And a voice can sound authentic without being real.
The mistake is looking only for visual flaws
There is popular advice about watching hands, eyes, shadows, lip sync or blinking. Sometimes these can reveal a manipulation.
But relying only on those flaws as a defense has a problem: the technology improves. A more robust strategy is verifying the context.
Ask first: where did it come from?
Not "does the video look real?", but "what is the original source?"
Check:
- who published it first;
- whether the same information exists in independent sources;
- whether the person in it actually posted it;
- whether reliable media are reporting the event;
- whether it only circulates through anonymous accounts or forwarded messages.
The more emotional it is, the slower you should react
Manipulated content works especially well when it triggers outrage, fear, surprise, excitement or hate.
A strong emotional reaction should be a signal to slow down, not to speed up.
Never make an important decision based on a video alone
If a supposed executive orders a transfer over video: verify through another channel.
If a relative appears on a video call asking for money: verify.
If a public figure seems to announce an investment: look for independent confirmation.
The best defense against deepfakes is not seeing better
It's verifying better.
AI can make a visual fake nearly perfect. But you can still ask:
- Who is showing me this?
- What do they want me to do?
- Can I verify it independently?
The book's artificial intelligence chapter covers deepfakes, voice cloning and AI-powered phishing.
DOWNLOAD A HACKER'S SECURITY GUIDE FREE →