The analysis, which examined more than 75,000 AI responses, found that only about one in five even included a news source. Among those that did, just a small group of publishers accounted for the vast majority of mentions. The BBC, The New York Times, and CNN made up almost a third of all references to media, with the BBC alone responsible for more than 11.37 percent.
Despite the study focusing on search queries in the United States, UK-based outlets like the BBC still came out on top. The findings show that only 12 news organisations received nearly 90 percent of all citations. In contrast, the other 18 included in the analysis were left to split the remaining 10 percent between them. The Financial Times, for instance, was cited 195 times less than the BBC for the same set of keywords.
Several other well-known names, such as MSNBC, Vice, and TechCrunch, also struggled to break through. Together, those brands appeared in less than one percent of all media mentions. Researchers say the issue lies in how Google’s AI system prioritises sources that are already well-established. It tends to rely on outlets with high authority and recognisable names, making it much harder for lesser-known sites to gain visibility—even if they cover the same stories.
What’s more, this pattern doesn’t always match traditional search results. Around 60 percent of the media links cited in AI Overviews didn’t appear in the top 10 organic search results for the same terms. That suggests Google’s AI doesn’t simply rely on page rankings but draws instead on a mix of perceived trustworthiness and content quality.
Inequality in citation was measured using a Gini coefficient, a common tool used to study economic inequality. The score came in at 0.54—signalling a significant imbalance in how attention is spread across news providers. A perfect score of zero would mean all sources were treated equally.
There are concerns too about how the AI handles content from paywalled sites. In many cases, long segments of text were copied from behind paywalls and included in AI Overviews. Of those responses that pulled from subscription-only material, nearly seven in ten contained copied phrases of five words or more. Some even went beyond ten words. Despite this, fewer than one in six of those examples gave proper credit, raising questions around fair use and licensing.
When AI Overviews do include news links, most of them are pushed to the bottom of the response. Fewer than two links are typically used, and over 90 percent are tucked into a links section rather than the main text. Media brands are also more likely to be linked than named outright. In fact, more than a quarter of media mentions appear without any clickable link at all, especially when the AI pulls information from aggregators instead of original reporting.
The type of search query also plays a part. People looking for news are more likely to see AI responses that include media citations, with news-related searches being more than twice as likely to contain media references compared with general ones. This opens the door slightly for outlets that specialise in breaking news or niche topics. Still, the broad trend heavily favours the larger players.
Although smaller publishers face clear challenges, the researchers point out a few ways they might improve their chances. Getting backlinks from sites already cited in AI Overviews could help. So could refining technical signals on their own websites—like properly using metadata to mark whether content is freely accessible. Publishers who focus on specific topics or underserved areas may also see better results over time, as Google’s AI seems to reward expertise in well-defined subject areas.
With Google’s AI Overviews now appearing in nearly one in five publisher-related searches, these patterns are likely to shape the future of online visibility. The old rules of SEO still apply, but increasingly, success depends on trust, authority, and depth of content—qualities that the AI appears to prioritise above all else.
Read next: Apple Rejects More Government Push Data Requests Amid Growing Demand
by Irfan Ahmad via Digital Information World
