AI Relies Primarily on Journalistic Content, Positioning the Press as Its Structural Foundation
Artificial intelligence relies predominantly on journalistic content, placing both specialized and general news media at the core of its informational foundation. According to a Cision report on media trends, 41% of the sources used by large language models (LLMs) to interpret CES 2026 come from technology media outlets, while 18% originate from general news organizations. This confirms that generative AI systems draw largely from journalistic content, making the press a structural basis for the responses delivered by systems such as ChatGPT, Google AI Mode, or Perplexity.
Large language models not only consume journalistic information; they reorganize it, synthesize it, and transform it into a new layer of informational intermediation that reshapes the visibility hierarchy of companies and sectors, according to the report “CES 2026: Media Trends Analysis” produced by Cision.
The study introduces an uncommon methodology in media coverage analysis. In addition to measuring the traditional media reach of CES 2026, it submits questions about the event to several generative AI systems — Google AI Mode, ChatGPT, Perplexity, and DeepSeek — and examines how they construct their answers. The goal is not to evaluate the trade show itself, but to understand how AI models interpret a major technology event and what role the information ecosystem plays in that interpretation.
A Structural Conclusion
The first conclusion is structural: the models do not reproduce the detailed specifics of product launches nor prioritize individual products. In their responses, robotics (28%) and video games (25%) emerge as dominant trends, ahead of sectors such as mobility, smart home technology, or display screens. The report emphasizes that GenAI systems tend to focus on forward-looking, strategic themes rather than specific announcements. This suggests that artificial intelligence reframes narratives around broad conceptual frameworks, pushing product-level granularity into the background.
Forty-one percent of the sources used by the models come from technology media, and 18% from general news outlets.
Journalism as the Core Source
The second conclusion directly concerns journalism. Forty-one percent of the sources used by the models originate from technology media, while 18% come from general news organizations. In other words, nearly 60% of the informational base behind AI-generated responses is journalistic in origin. The official CES website accounts for 15%, corporate websites 9%, and YouTube 6%.
The report underscores that journalistic content remains the primary source for LLMs, placing media organizations at the center of the AI training and consultation ecosystem. This finding carries important implications for the information industry: although models synthesize and reformulate content, they depend on the prior production of both specialized and general media outlets.
AI does not displace journalism as a primary source; rather, it functions as a layer of aggregation and reinterpretation. Editorial authority and analytical depth remain decisive in determining whether a narrative is absorbed by generative systems.
Diverging Visibility: Media vs. AI
The study also identifies a divergence between traditional media visibility and visibility within AI-generated responses. Nvidia leads both in media reach (20%) and in LLM mentions (8%). However, other companies display different patterns. Samsung, for instance, achieves proportionally higher performance within language models due to the strong presence of its own website as a cited source. This indicates that a company’s digital architecture directly influences how AI systems represent its brand.
In addition, companies such as Siemens, Google, XREAL, and ROG appear prominently in AI-generated responses despite having lower rankings in traditional media coverage. According to the report, this is explained by the frequency with which these brands are mentioned in indexed sources consulted by the models. Algorithmic hierarchy does not necessarily align with editorial hierarchy.
A Dual Competition for Visibility
From a corporate communications perspective, the report points to a shift toward a dual competition for visibility: one media-driven and the other algorithmic. The former depends on press coverage and positioning within the news agenda; the latter on digital citability, semantic content quality, and the ability to integrate into the corpus used by language models.
Individual leadership is also amplified by this dynamic. Jensen Huang accounted for 51% of the visibility among the executives analyzed in media coverage, and his strategic messaging around physical AI and infrastructure was interpreted more as a forward-looking vision than as a product presentation. This strategic orientation aligns with the type of narrative LLMs tend to prioritize, reinforcing the relevance of conceptual messaging over tactical announcements.
The report concludes that public influence is no longer defined solely by the volume of press coverage, but also by how that coverage is absorbed, indexed, and reformulated by artificial intelligence systems. For media organizations, the findings confirm their role as the structural source of the digital memory that feeds AI. For brands, they introduce an additional variable: appearing on the news agenda is no longer sufficient; what matters increasingly is how that presence translates into citations within generative environments that are becoming central intermediaries in access to information.
Source: Laboratorio del Periodismo
Author: Redacción
Picture from Freepik