- AI chatbots appear to be providing less diverse responses, research claims
- Displaying the same answers has been likened to globalisation, where the same products and brands are available everywhere
- Such a move may increase convenience, but also reduce diversity
Could we be heading into a real-world Idiocracy? A repeated reliance on AI to provide answers – invariably the same answers, offering the same views – could be shrinking human knowledge, a new study has claimed.
The research, which tested 27 different LLMs, found that even with the best performing LLM, the information presented as answers was almost 19% less diverse than what is available using Google’s standard search engine. Consequently, the information was less varied, providing a narrower selection of answers, views, and restricting interpretation.
Researchers believe that this trend risks pushing certain content to the peripheries of accepted knowledge, while other content is considered dominant or primary – and the end result of this could indeed be a world where people are less aware, more idiotic, and perhaps even with less intelligence than today.
Topic testing
The study was conducted by researchers at the Department of Computer Science (DIKU) at the at the University of Copenhagen.
To establish how the LLMs would deliver answers, they were tested on 155 topics (ranging from nuclear weapons to K-Pop). The researchers used 200 prompt formulations drawn from research into real AI chatbot usage, generating an incredible 70 million responses from the models.
Even answers from GPT-5, the LLM producing the most varied responses, were still 18.5% less varied than answers from Google Search.
“People are to a large extent exposed to the same information over and over again,” says first author Dustin Wright. “So AI chatbots are not just changing how we find knowledge, but also which knowledge we have access to."
An Idiocracy?
Whether using AI models as part of Google Search, DuckDuckGo’s search tool, Bing, or by specifically quizzing Gemini, ChatGPT, or any of the other chatbot-accessible LLMs, the findings from the research indicate a change in how information is surfaced.
“We risk exposing people to fewer perspectives and a narrower range of knowledge. This could create a vicious cycle in which the most popular content becomes even more dominant, while other content is increasingly overlooked,” says DIKU Professor (and senior author) Isabelle Augenstein.
She compares the phenomenon with the downsides of globalization: “Today, you can buy the same products and find the same coffee chains almost everywhere in the world. That has many advantages, but it has also reduced diversity.”
As for the possibility that AI could shrink knowledge across humanity, and leave a world governed by those wholly unsuited to the task, Professor Augenstein is more upbeat. She says that a “knowledge collapse is not happening yet […] we can see that the more recent models produce slightly more diverse answers than older models.”
