Oh No, AI Journalists Are Actually Covering Major Stories!

This approach carries clear risks: He places his trust in machines to evaluate what is factual, newsworthy, and what could lead to legal issues. One of his agents analyzes the legal risks associated with a story and assigns it a score; he refrains from publishing anything that is deemed overly risky.

He isn’t alone in this endeavor. Dakota Carrasco, a portfolio analyst at BlackRock, operates a venture called The Dissent in his spare time, functioning as an “agentic newsroom.” Similar to RuntimeWire, The Dissent is a solo operation supported by numerous bots and operates on a tight budget. According to Carrasco, the main site, focused on San Francisco, costs less than $1,000 per month to maintain. In contrast to Merket, who credits himself as the author of every RuntimeWire piece, Carrasco keeps his identity hidden while his “newsroom” operates.

Since its launch in March, he has developed various personas for his synthetic journalists. For instance, Bex Connolly, a City Hall beat reporter, is characterized as “skeptical without being snide,” while “sports degenerate” Sal Moreno provides Giants updates with “no bro-science, no Rogan-like credulity, no right-wing grift.” His focus is on aggregation, yet it doesn’t strictly adhere to citation norms—the bot reporters typically mention their sources but do not provide hyperlinks. (“I’m working on improving that,” Carrasco assures.)

Nicholas Diakopoulos, a professor at Northwestern University and director of the Computational Journalism Lab, regards this as an “experimental phase” for media startups leveraging generative AI tools. “It’s unclear to me if there’s a significant audience for these AI-written news platforms,” he comments. He also expresses skepticism regarding mainstream journalists, who prefer to control the language and framing of their stories to ensure integrity, legality, and accuracy, willingly relinquishing control to AI.

Diakopoulos has noticed that when AI chatbots search for sources, they often retrieve AI-generated articles. In an upcoming paper, he and a colleague discovered that AI tools like ChatGPT and Claude cited AI-generated sources 16 percent of the time across four different topics. Diakopoulos speculates that this tendency for AI to source synthetic writing might help AI newsrooms attract readers: “That could be one avenue through which some of this content reaches a human audience.”

Pete Pachal, founder of a newsletter and podcast focused on generative AI and media, expresses concerns about the ability of an AI newsroom to produce certain types of reporting that depend on traditional sourcing. “I simply don’t foresee that happening,” he states. “Building the trust of sources is something I believe will remain a human-only task.” However, he sees potential for these projects in specific areas of journalism, especially in extracting scoops from large datasets or covering live events like an Apple product launch, viewing them as a “natural evolution” in the utilization of these tools. “Honestly, it feels a bit unavoidable,” he adds.

Is this form of reporting genuinely journalism? “I strive to adhere to journalistic ethics and standards,” Merket states. He mentions contacting companies and individuals referenced in his articles for comments before publishing, linking to sources when aggregating news, and providing corrections when necessary (so far, he has issued three). Sometimes, he adopts a reporter’s tone: “This weekend, I was thrilled to share two exciting scoops.” Other times, his demeanor leans more towards the Silicon Valley mentality. He recounted a situation where his AI agents uncovered actual scoops about startups by sifting through company websites and how he later retracted those stories after the companies involved requested it—not due to inaccuracies, but as a courtesy. “Founder to founder, it’s understandable,” Merket reflects.