Lesson 3 of 6

Spotting fakes

6 min read

You've heard the tricks: count the fingers, check the text, look for the too-smooth skin. Problem is, the models fixed all that — and the detectors meant to catch them keep getting it wrong.

Your eyes stopped working

For a while you could spot fakes by their mistakes. Models improved fast, and those tells are disappearing — a convincing AI image, voice, or video is now cheap to make. Trusting your eyes is exactly the instinct that gets people fooled by a good fake.

You can't reliably eyeball AI media anymore — and detectors only guess.

Detectors guess, and miss

So people reach for an "AI detector". But these tools output a probability, not proof, and they fail both ways: they flag real photos as fake (a false positive) and wave real fakes through as genuine (a false negative). Betting a reputation — or an accusation — on that number is a mistake.

A detector's score is a guess, wrong often enough that a verdict is never proof.

Never treat an "AI detector" result as proof, in either direction. A real photo can be flagged as fake, and a good fake can pass as real.