A study by Carnegie Mellon mapped the ownership of 9,761 digital media in 77 countries and measured the effect of seven AI systems on information concentration.
When a search engine responds with a generated summary, such as Google's AI Overviews, it decides which media outlets are directed to for attention and revenue. To measure this effect, Peter Carragher and Kathleen Carley (Carnegie Mellon) used Wikipedia and Wikidata to compile a database of owners of 9,761 digital media outlets, validated against independent U.S. and U.K. directories, and derived audience-weighted concentration indices for 77 countries.
More concentrated than a merger under surveillance
The first finding is that journalists report being more pressured by their owners where concentration is high, and the headline metrics underestimate it, with those calculated per owner being 1.3 times higher. Applied to the seven AI systems studied, these metrics show citations up to 2.08 times more concentrated than actual traffic. If AI Overviews were used as an intermediary for all organic search in the U.S., the increase in concentration would be several times higher than that associated with the most tightly regulated media mergers.
Another worrying result is that when queried in a country's national language, AI systems prefer to cite state-controlled media. The authors publish their dataset to enable global monitoring of the impact of disruptive technologies on media pluralism.
Read the full article on the original source
Go to the source →