GPT Guesses Between 1 and 100

GPT-4.1 produces decidedly non-random distributions when asked to pick a number between 1 and 100, exhibiting distinctly human biases rather than uniform randomness. The model shows extreme spikes at numbers like 37, 42, and 73—classic human preferences—and completely avoids multiples of 10, mirroring human number-picking patterns so closely that a chi-square test shows statistically significant deviation from a fair distribution. The findings suggest that language models inherit the statistical quirks of their training data rather than generating truly random outputs.