AI chatbots are slowly choking humanity as research shows a 'vicious cycle' could cause global knowledge collapse

A recent study by researchers at the University of Copenhagen's Department of Computer Science has found that AI chatbots are providing less diverse responses, potentially leading to a global collapse of knowledge. The researchers tested 27 different large language models (LLMs) and discovered that even the best-performing model presented information that was nearly 19% less diverse than what is available from Google's standard search engine.

This lack of diversity could narrow the range of answers and views available to users, restricting interpretation and potentially pushing certain content to the sidelines, while making other content dominant. The study, conducted by Dustin Wright and Isabelle Augenstein, involved testing the LLMs on 155 topics and generating 70 million responses using 200 different prompts.

Even the most varied LLM, GPT-5, was found to be 18.5% less varied than Google Search. Professor Augenstein compares the situation to globalization, where the same products and brands are available everywhere, which has increased convenience but also reduced diversity. She warns that a "knowledge collapse" is not imminent, but the trend is concerning, as more recent models do produce slightly more diverse answers than older ones.