Is it a butterfly or a bear? It all depends on how you rotate the image. Thanks to AI, scientists can now generate hundreds of such optical illusions, where a single image contains multiple recognizable figures depending on its orientation.

These “visual anagrams,” developed by Tal Boger and Chaz Firestone from Johns Hopkins University, help researchers study how the brain recognizes concepts such as size, shape, and animals — without being influenced by color, texture, or other visual features. According to Rob van Lier from the Donders Institute, these images are a powerful tool for precisely measuring perception and brain responses.