Study says AI chatbots give more uniform answers than web search

In short: A new academic study says AI chatbots, including Claude, tend to give a narrower and more similar set of answers than traditional web search.

What happened

Researchers led by the University of Copenhagen compared answers from several AI chatbots with results from Google-style web search. The reporting says the study found that every language model tested gave more uniform information than web search across the topics examined.

A “language model” is the kind of AI behind chatbots that write responses in plain text. You can think of it like asking one very talkative assistant for a summary, instead of looking at many different sources yourself.

The article also reports that smaller AI models gave a wider mix of information than larger ones. Newer models were more varied than older models. Even so, all of them still produced less diverse information than standard search engines.

One detail in the reporting involves Anthropic’s chatbot Claude. An expert at the University of Copenhagen said the team had been sharing its research with Claude, and that the model’s new output matched what the researchers found. The available summary does not include the researcher’s name or a full quote, and it does not mean Claude independently verified the study.

The study has reportedly been accepted to EMNLP 2026, a major research conference scheduled for October 2026.

Why it matters

Many people now use chatbots as a shortcut to learn about health, politics, and everyday decisions. If chatbot answers are more uniform, it can be easier to miss minority views, local context, or important disagreements. Web search can act more like a bookshelf with many authors, while a chatbot can act more like a single, confident narrator.

Source: NYTimes