Sign up for the daily CJR newsletter.
Since it was first published, in 2001, The Elements of Journalism, by Tom Rosenstiel and Bill Kovach, has schooled fledgling reporters in their craft by advocating that journalists adhere to a series of core principles—including that journalism’s first obligation is to the truth; that it is, in essence, a discipline of verification; and that journalists must serve as a monitor of power. In Rosenstiel’s new book, The Next Journalism: How the Press Must Change to Serve Democracy, released last week by Crown, he argues that his intention is to deepen these principles, not to replace them. But his new recommendations tilt in a different direction. Rosenstiel, a professor at the University of Maryland’s Philip Merrill College of Journalism, argues that journalism is “failing democracy” and should “become more focused on helping people live their lives than on merely getting their attention.”
To achieve this aim, he suggests we reframe our traditional understanding of newsgathering, such that professional journalists, the public, and “machines” all have “distinct, essential, and complementary—not competing—roles.” Notably, he writes that “journalism must run toward Artificial Intelligence, not away from it, to make journalism stronger.” As he explained to me last week: “We should be using AI in journalism to make our journalism better, not primarily to make it cheaper.” And, he acknowledged, “it will be a point of friction in the book.”
But what exactly would it look like to “run toward” AI? One of his suggestions in The Next Journalism is that news organizations, with the help of AI systems, serve their audiences by assembling “the vast array of public data about community life into a new and easily usable Civic Internet.” In other words, instead of using datasets for one-off enterprise stories, journalists should work to combine multiple streams of information into a single, constantly updating hub—taking on the role of data wranglers, one might say—and then enable audiences to navigate that resource themselves, for instance with an AI chatbot. This, he argues, could help put newsrooms back at the center of their communities. The idea is about “extending the reach of human journalism using technology as a tool,” he told me. “I’m not advocating having AI write stories.”
Rosenstiel gives other use cases for AI systems: auditing reporting to ensure under-covered communities are well-represented; identifying false claims from politicians that could then be fact-checked; collecting and analyzing data about audience needs and preferences. “AI is going to happen. It’s the next digital revolution,” Rosenstiel told me. He acknowledges problems around hallucinations, bias, environmental concerns, and so on. Even so, he argued, models are getting better all the time. And: “If we say, ‘AI is bad, I don’t want to use it,’ it will happen to us.” Without serious journalists getting involved, he said, bad practitioners will move into the space, promising news without the professional ethics or standards of serious journalistic organizations. “That’s why we have to run at it, and run at it with the concept of: How do you make it better?”
Some news organizations are headed in that direction. The Washington Post (Ask the Post AI) and Financial Times (Ask FT) have made chatbots from their journalism, and Jim VandeHei, the cofounder and CEO of Axios, talked to Issie Lapowsky for a recent CJR story about Axios’s partnership with OpenAI and what he views as its potential to “save local news.” Many big publishers are cutting licensing deals with tech firms.
But the idea of embracing AI is harder to swallow for others. In March, Emily Bell wrote for CJR about how a group of news organizations—some from opposite ends of the political spectrum, like The Guardian and Telegraph—have formed the Standards for Publisher Usage Rights, or SPUR, to advocate for guardrails around responsible AI use, among other things. It was encouraging to see an American company, the Associated Press, join SPUR in July, particularly in the absence of political will in the United States for a conversation about meaningful regulation (as in the European Union’s landmark AI Act). Meanwhile, where journalists feel their companies are not taking the troubling aspects of AI seriously enough, as Riddhi Setty reported for CJR in April, unionized employees are fighting for contracts that set terms for its use.
Newsroom opposition to hasty AI adoption comes amid high-profile inaccuracies and citation problems from large language models (LLMs)—which surely break Rosenstiel and Kovach’s rule about journalism’s first obligation being to the truth—as well as a litany of disputes with tech companies over copyright infringement. (These are helpfully collated in the Tow Center’s AI Deals and Disputes Tracker.) “I do not understand this idea that journalism should rush toward AI, particularly in the editorial process,” Courtney C. Radsch, the director of the Center for Media and Digital Governance at the Open Markets Institute, told me. She pointed out that “now you have information operations targeting LLMs” from state actors such as Russia, as Wired has reported, as well as “ideological biases embedded in the LLM,” the impacts of which are difficult to identify. “I mean, sure, do it in your admin functions,” Radsch said. “But the idea that we can use it in the creation and production of journalism, I think, is very problematic—because then what differentiates journalism from anyone else who uses an LLM to create content?”
Felix M. Simon, a research fellow in AI and digital news at the Reuters Institute for the Study of Journalism, found, in a report for Aspen Digital in March of last year, that there is now a premium on distinctive, human-made reporting that cannot easily be made or replicated by AI systems, such as original investigative reporting, nuanced analysis, and stories that come from long-term relationships with communities. Simon told me that some publishers are now thinking, “We have to provide something that is unique, that people cannot easily get elsewhere, and that people are willing to pay for.”
Simon also looked at how European news organizations are approaching the implementation of AI systems, concluding that there is “incremental” adoption rather than “revolutionary transformation.” (The report grew out of a gathering of about eighty UK and European media leaders who discussed their companies’ approaches to AI.) Most news organizations have focused on efficiency, the report found, and “on automating routine tasks like transcription, translation, and headline creation.” Simon characterizes their attitude toward AI as “a mixture of caution and excitement, with the latter seemingly more pronounced at the senior leadership level.”
Overall, Rosenstiel’s book contains interesting ideas about equipping journalists for the twenty-first century. He calls on news organizations to better understand the role their reporting plays in the daily lives of the communities they cover, as well as to collaborate much more intentionally with audiences. That is all very welcome. But the most sensible approach to AI tools for news organizations may be the one they’re already taking—not running toward shiny objects but walking around them, cautiously, and verifying whether they’re worth it.
This piece was produced with support from the Craig Newmark Center for Journalism Ethics and Security.
Has America ever needed a media defender more than now? Help us by joining CJR today.
