Journalistic content represents a significant portion of the responses generated by AI models, with an overall citation rate of 27%. This number increases to 49% for queries requiring recent information. The study revealed that when citations are disabled, generative AI provides less reliable outputs. Variations in citation sources were noted based on the question type, such as subjective questions leading to greater citations of corporate blogs. Notably, different models exhibited differing tendencies in citing journalistic sources, with Claude citing them the least frequently.
Journalistic content was cited more than 27% of the time across all Muck Rack's tests, rising to 49% for queries needing recency, such as timely news.
When citations were disabled for the generative AI models, they produced unreliable information, lacking access to real-time data, confirming a dependence on journalistic sources.
Subjective queries led generative AI models to rely more heavily on corporate blogs and content, demonstrating a shift in citation sources based on the question type.
Claude cited journalistic sources less frequently compared to other models, utilizing Reuters far less often, which highlights variations in citation behavior among AIs.
Collection
[
|
...
]