A new study led by the University of Copenhagen has warned that heavy reliance on artificial intelligence (AI) chatbots for information can trigger a “knowledge collapse”, as users are exposed to a far narrower range of facts and perspectives than through a basic web search.
Replacing Google, AI chatbots have become a common way of getting answers to everything from what to make for dinner and how to word that difficult email to the boss, to what it actually means when interest rates rise, the study noted. An AI chatbot is a service such as ChatGPT, Gemini, or Claude that one can communicate with by typing or speaking and that provides answers in return. The chatbot uses a so-called large language model to understand questions and formulate responses, the report mentioned. Across all topics, every chatbot delivered more uniform responses than a simple Google search, meaning users repeatedly encountered the same information and viewpoints. “Every language model we tested provides users with more uniform information than a simple Google search across all the topics we looked at. In other words, people are to a large extent exposed to the same information over and over again. Smaller models sometimes produced more varied answers than larger ones, and newer models were better than older ones, but none matched the diversity of a search engine that links to many websites. “We risk exposing people to fewer perspectives and a narrower range of knowledge. Researchers have warned that if future AI systems are trained partly on text generated by other AIs, which is already less varied, each generation could become progressively narrower, gradually shrinking the range of information available through chatbots. Augenstein stressed that AI chatbots should not be treated as the only source of information. For students and users, practical steps include: AI chatbots are powerful aids for drafting, summarising and quick revision, but the Copenhagen study has shown that they currently offer a narrower knowledge diet than traditional search.
Researchers at Copenhagen’s Department of Computer Science (DIKU) tested 27 large language models across 155 topics linked to 12 countries, generating about 1.7 million answers and 70 million claims. Even the most diverse model tested, OpenAI’s GPT-5, offered at least 18.7 per cent less varied information than Google.
So AI chatbots are not just changing how we find knowledge, but also which knowledge we have access to,” says first author Dustin Wright, a former postdoctoral researcher at DIKU who is now an assistant professor at Aalborg University. The study defined “narrower slice of knowledge” as reduced variety in content, not shorter answers: chatbots tend to repeat similar claims and angles rather than surfacing multiple sources, contexts or interpretations. This could create a vicious cycle in which the most popular content becomes even more dominant, while other content is increasingly overlooked,” a professor at DIKU and senior author Isabelle Augenstein, said. This self-reinforcing cycle is what the team calls “knowledge collapse”.
