How to tell an AI-generated photo from a real one: 5 signs that still work

A fake screenshot of "a celebrity endorsing this token," or a photo of a fake "exchange representative" giving away prizes, is one of the most common tools in a crypto scammer's kit. AI image generators have gotten noticeably better, but a few technical weak spots still give fakes away.

1. Hands and fingers

Even modern image models most often slip up on hands: an extra or missing finger, an unnatural joint bend, or a strange hand position relative to whatever it's holding. This remains one of the most reliable tells.

2. Background text

Signs, text on clothing, text on screens in frame — AI models frequently generate something letter-shaped that doesn't actually spell real words, especially if you look closely at background details rather than the center of the frame.

3. Inconsistent reflections and shadows

Check whether shadow directions match the obvious light source in the photo, and whether reflections in mirrors, windows, or glasses match what should actually be reflected. AI models generate each part of an image based on local plausibility, so overall physical consistency in a scene sometimes suffers.

4. Overly "smooth" skin and texture

A lack of pores, asymmetry, or minor skin/hair irregularities — especially paired with perfect lighting across the entire frame — is a typical generation side effect, though the newest models are gradually getting better at faking this too.

5. Reverse search and metadata checks

A practical, non-visual step: a reverse image search will show whether a similar photo was published before in a different context, and the absence of camera EXIF metadata (device model, shooting settings) on a supposedly "live" photo is another reason to be suspicious.

What this means in practice

No single sign is foolproof — models improve fast, and some of these signals may become less reliable within a year. But a combination of a few one- or two-minute checks still catches the vast majority of fake "proof" used in crypto scams.

This material is for educational purposes only.