One in four quotes in AI chatbot responses comes from journalism, Muckrack study finds

Editor's take

An analysis of 15 million AI citations reveals that journalism constitutes a quarter of the sourced material for large language models like ChatGPT, Claude, and Gemini. This reliance highlights the enduring value of curated, factual reporting in training AI, particularly benefiting niche publications and specialized reporters over broader news outlets.

The finding underscores the complex relationship between AI development and the established media ecosystem. It raises questions about fair compensation for content creators whose work forms the bedrock of these generative models, and the potential for future AI training data to become even more concentrated in specific journalistic domains.

Future developments to monitor include potential licensing agreements between AI companies and news organizations, and whether AI developers begin to diversify their training data sources beyond traditional journalism to mitigate bias and ensure broader representational accuracy. The Muckrack study's continued analysis will be crucial in tracking these evolving dependencies.