Detecting Psychopathological Visual Patterns in AI-Generated Art: A Neural Network Approach to Identifying Schizophrenic-Style Signatures

We trained a fastai/PyTorch-based neural network to distinguish between four classes of images drawn by healthy and mentally ill people. A comparative evaluation involving Gemini, ChatGPT, Grok, and Qwen revealed that AI simulates psychopathological markers with such accuracy that our network classifies them as “schizophrenic” with 75% confidence . The model is available as a Telegram bot for open testing.