A Simple Technique to Identify Whether an Image Was Generated by AI
A technique has finally been discovered that anyone can use to identify whether an image was generated by AI or not. While in most cases it’s obvious, there are still illustrations where it’s hard to tell if AI was involved.
The method went viral thanks to a user called “Fuck Ai” (yes, really), whose post explaining the technique reached nearly 6 million views. The process works as follows.
Take any illustration you suspect was created using AI, preferably one with a white background. Then, using an image editing program, increase the contrast and saturation. If the image was generated by AI, it will look like this.
If it was created by a human, the image will tend to look like this.
The logic behind this is simple:
- In a 100% digital illustration, a pure white background usually remains white, without color variations.
- In images generated or modified by AI, it’s common for micro color variations and noise to appear, even in areas that should be completely white.
- When contrast and saturation are increased, this hidden noise becomes visible.
- According to the original post, this effect does not come from image compression or uploading, and it can be reproduced as a test.
So there you have it: a way to identify whether an image may have been generated by AI or not. Even if the background appears to be a single color, AI-generated images often contain hidden artifacts.
Some people were very happy with this discovery, while others argued that the method is not perfect. One user even shared an example of a very noisy image that the artist insists was made by a human.
So, what do you think of this technique?