Research Editor selected
LLM-based search surfaces less local, less diverse news than search engines in Latin America
A new arXiv study compares three search engines and four language models across 1,901 current events in nine Latin American countries, varying query language and user location. Across 49,207 requests and 8,834 unique domains, language models returned a less diverse set of news sources than search engines. For local users, search engines already returned only 41.3% local news, and language models returned 3.6 to 11.6 percentage points less. The authors release an open dataset to support multilingual and geographically diverse evaluation of LLM-based search.
Why it matters: It quantifies how LLM-based search can reproduce or amplify geographic bias in news visibility, with an open dataset for evaluation.
LLM-based searchnews diversityLatin America