Publishers are wondering how generative systems source their content.
The answer is that non-paid media accounts for 94% of all links cited by AI, according to a recent study by Muck
Rack. This includes journalistic sources (24.7%) and third-party corporate blogs (24.5%). The only “paid” channel is press releases (6%).
Corporate blogs qualify under
“earned media.” However, the earned category has declined to 82% from 89% in July.
The cause? The “reliance on aggregator/encyclopedic sources decreased
during this period as OpenAI lowered their reliance on Wikipedia,” the study states.
Muck Rack analyzed 1 million+ links from AI responses.
To put this in
context, Muck Rack is specifically interested in public relations. It found a mere 2% overlap between the journalists most pitched by brands and those cited most by AI engines.
“This insight means an adjustment to pitching practices could have major impact on a brand’s GEO [Generative Engine Optimization],” Muck Rack writes.
advertisement
advertisement
Here are some
additional insights for publishers trying to navigate this ecosystem, from the executive summary:
- Reddit is one of the most critical domains. And 51% of the answers citing it also cite
a journalistic source.
- AI prefers recent sources--over half of all citations were published in the last year. And those published within the last seven days are most
successful.
- AI models consider the authority of a source. “This can be in a broad sense, like high domain authority sites like Reuters,” the study states. “Or in a
more targeted way, like citing
industry-specific sources for industry-specific queries.”
(Note: In other news: Nancy Barnes is leaving her post as editor in chief
of The Boston Globe, effective this week. Barnes, the first female editor in chief of the Globe, will continue to work as editor-at-large on the publication. She hopes to step back
from the daily grind. No replacement has been named).