AI Chatbots Overwhelmingly Pick 27 When Asked for a Random Number

A data scientist named Mohd Faraaz found in June 2025 that six out of seven major chatbots — including ChatGPT, Gemini, and Claude — all chose 27 when asked to pick a number between 1 and 50. A follow-up study by Javier Coronado-Blázquez in Madrid tested six models across seven languages and multiple number ranges a total of 75,600 times, confirming that every model showed strong numerical biases. The root cause lies in how large language models process information: to an LLM, numbers are simply text tokens with no inherent mathematical value, making truly random selection impossible. Interestingly, humans are no better — a 1976 study found people gravitate toward 7, and a Veritasium survey of 200,000 respondents showed 7, 37, 73, and 77 as top picks. Researchers note this is not a bug to be patched but a fundamental characteristic of how both AI models and human minds work.