The nature of the large language models underpinning AI chatbots is such that they give us a significantly narrower range of information than a conventional Google search, researchers say.

Instead of relying on traditional Google searches, a growing number of people now turn directly to AI chatbots for answers.

Chatbots have become a common way to get 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.

Less different information

However, researchers at the Department of Computer Science at the University of Copenhagen (DIKU) warn that because of how large language models function, they tend to provide a much more restricted range of information.

“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,” says first author Dustin Wright, a former postdoctoral researcher at DIKU who is now an assistant professor at Aalborg University.

“So AI chatbots are not just changing how we find knowledge, but also which knowledge we have access to.”

Wright and his colleagues, testing 27 different large language models on 155 topics, generated a dataset of around 70 million individual claims produced by the models.

The topics tested ranged from nuclear weapons, marriage, pornography, racism, and genocide to more country-specific topics such as Marine Le Pen, the Falklands War, and K-pop.

The results of the study show that even the language model producing the most diverse answers – OpenAI’s GPT-5 – provides at least 18.7 percent less varied information than Google.

And since people increasingly use AI chatbots as their gateway to information, consequences could be significant, Professor at DFIKU and senior author Isabelle Augenstein said.

Knowledge collapse on the horizon

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,

Professor at DFIKU and senior author Isabelle Augenstein said.

The reason data diversity is so low is the way models work.

They compress huge amounts of text they’re trained on and learn the patterns that occur most frequently – but in the process, information that deviates from the most common patterns is filtered out.

The effect could be amplified if language models are increasingly trained on text produced by other AI models – something the researchers expect to happen.

In that case, models would learn from their own outputs, which are already less diverse than the human-written texts on which they were originally trained.

If this process is repeated over several generations of models, the range of information could gradually become narrower. This is what the researchers refer to as ‘knowledge collapse’.

“It’s a worrying thought. We can see that the more recent models produce slightly more diverse answers than older models, so knowledge collapse is not happening yet,” said Augenstein.

“But the mechanism that could trigger it in the longer term is already there. So it is something we should be aware of.”