Plus: U.K. journalists’ skeptical adoption of AI, politicians as drivers of journalist harassment, and metrics’ post-publication influence

Editor’s note: Longtime Nieman Lab readers know the bylines of Mark Coddington and Seth Lewis. Five years ago, they launched a monthly newsletter on recent academic research around journalism and we’re happy to bring each issue to you here at Nieman Lab. In the last couple of months, they have added two other journalists-turned-academics to their team: Tamar Wilner and Nick Mathews, who teach and research at the University of Kansas and the University of Missouri, respectively. This month’s edition was written by Nick and Mark.

Henry Jenkins famously wrote, “If it doesn’t spread, it’s dead.” With that concise observation, Jenkins captured a central truth of the digital era. Information, news, and media gain power when people choose to circulate them. Sharing amplifies the influence of news organizations, extending their reach across networks far beyond their own platforms. That insight has shaped much of what we understand about news sharing on social media and explains why news outlets invest heavily in optimizing content for these spaces — even as they continue to struggle to convert visibility into financial sustainability.

But focusing on what spreads tells only part of the story. Social media may be a major vehicle for encountering news, yet people publicly share only a small fraction of what they consume. Most news remains private — read but never posted, forwarded, or otherwise made visible. While research has examined virality extensively, the equally consequential behavior of deliberately withholding news has been largely overlooked. Withholding is not merely an absence of action. It is an active choice that shapes which stories enter public discourse and which remain private.

In a new study published in Social Media + Society, scholars Jennifer Ihm and Eun-mee Kim explored this understudied space to understand not just whether people withhold news, but why. Drawing on surveys from more than 400 social media users and a textual analysis of the news items they chose not to share in their three most active KakaoTalk chatrooms — Korea’s dominant messaging platform — the researchers developed a model of deliberate news withholding grounded in self-presentation. Their findings show that sharing is rarely a simple yes-or-no decision. Users carefully navigate a nuanced process, withholding news for specific self-presentation reasons and tailoring those choices to particular audiences. They don’t avoid sharing everything with everyone. They strategically withhold certain stories in different networks, shaping how they are perceived and managing their online identity.

To explore these dynamics, the researchers combined survey data with detailed text analysis. Survey responses revealed three key self-presentation goals — shaping one’s own identity, protecting privacy, and managing or pleasing the audience. News was categorized by type — hard or soft — and by topic, including politics, economics, society, lifestyle, world news, and science/technology. Each story was analyzed for emotional content — happiness, fear or sadness — and argumentative content, reflecting how strongly it sought to persuade or refute a point. Finally, social context was measured by the size of each user’s top three chatrooms and the strength of ties between members, based on frequency of group conversations.

The findings reveal that different user-audience networks activate distinct self-presentation goals. In smaller, closer networks, users are more attentive to pleasing their audience and maintaining relationships, carefully weighing the potential impact of each story. In larger, more public-like networks, these concerns are less pronounced, as it becomes impossible to satisfy every participant. Tie strength also plays a role. Weaker connections make users more cautious about revealing private or identity-related information, while closer relationships prompt greater attention and deliberation. Put simply, the closer the relationship, the more carefully users consider which news to share or withhold, tailoring their choices to both content and context.

The type of news also interacts with network size in meaningful ways. Users are especially likely to withhold highly argumentative news in larger networks, where such stories are perceived as riskier. Network size thus provides insight into the communication context. Small, tight-knit networks feel private and manageable, whereas very large chatrooms resemble public communication, requiring users to carefully calibrate what they share and what they withhold.

Together, these findings reveal how the interplay of content, audience, and social context shapes deliberate decisions to share or withhold news, offering a more complete understanding of online discourse. Ordinary social media users are not passive recipients. They actively evaluate both stories and audiences before deciding whether to amplify or suppress information. They navigate nuanced relationships and balance self-presentation goals with concerns about emotional impact, argumentativeness, and audience size. Echoing previous research, the authors demonstrate that users act as gatekeepers in the distribution of news, shaping what becomes visible and what remains private.

For journalists and news organizations, these findings offer practical lessons. First, social media audiences are far from uniform. A story that resonates in one network may be ignored — or deliberately withheld — in another, depending on audience expectations, norms, and the relationships among members. News sharing is inherently relational, shaped by how audiences interpret content and manage their own social identities, which means journalists can influence but never fully control how a story spreads. This makes it essential to consider not only what is reported but also how it is presented, including tone, framing, and emotional appeal, as subtle differences can determine whether a story is amplified or quietly consumed.

Second, engagement metrics like likes, shares, and comments capture only a fraction of a story’s true reach. Much news is read silently, without visible signals of attention. To understand impact more fully, newsrooms should complement social metrics with surveys, polls, or other forms of audience feedback, gaining knowledge into both who is seeing the content and how it is being interpreted or valued.

Research roundup

Amid the continual clamor about the use of AI in journalism, it can be difficult to get a timely and comprehensive picture of how journalists are actually using AI in the newsroom. Anecdotal descriptions and examples from individual news organizations have value and are easier to come by, and numerous surveys have been published on journalists’ attitudes toward AI, but the overall picture of what journalists are doing with AI remains hazy.

That’s why the Reuters Institute’s new study, published last week, is so valuable. Through a survey conducted last fall with 1,004 U.K. journalists, we get a detailed snapshot of journalists’ use of AI and their news organizations’ approach to it throughout one country.

The study’s lead finding may be that more than half (56%) of U.K. journalists use AI in their work at least weekly, while just 16% have never used it in their journalistic work. The most common tasks (transcription, captioning, and translation) are less headline-grabbing, but more substantive journalistic tasks like story research, idea generation, and headline generation are fairly common too, with about a quarter of journalists saying they’ve done it at some point. Younger journalists are more likely to use AI professionally, but so are journalists with more management responsibility.

Across almost every group, U.K. journalists hold generally negative views of AI, being much more likely to see it as a significant threat to journalism than a significant opportunity. Those in upper management are far more likely than others to see it as a major opportunity — 34% — though 58% of them still view it as a significant threat. Journalists’ biggest ethical concerns about AI were in its potentially negative impact on public trust in journalism and on the journalistic value of accuracy and originality of content.

Despite this fairly widespread use, journalists reported relatively little integration of AI into their news organizations as a whole. Eighty-five percent of journalists said it was either not integrated at all or had limited integration into their organization. Few, though (20%), saw their organizations as opposed to AI. About 60% of journalists said their organization had guidelines for at least some part of AI use, though this was much more common in public media than commercial media. Public media and media conglomerates are also more likely to provide AI training, though only 32% of journalists had training available overall. Perhaps we might summarize U.K. news organizations’ approach to AI as cautious managerial optimism, applied rather haphazardly among a more skeptical newsroom.

The question of how journalists determine which stories are most important has been propelling a strain of research for at least 75 years, when a theory known as gatekeeping was first applied to news. For the past decade or so, scholars have been zeroing in on the role of analytics in the process, finding that stories that are likely to generate big traffic numbers are more likely to be covered and to get more prominent homepage placement, and less likely to be demoted or removed.

Riemann and Ehrlich’s study builds on this work, looking at how analytics and a story’s topic factor into journalists’ decisions on whether to remove a story from the homepage. They conducted a survey-based experiment with 412 journalists across Germany, Austria, and Switzerland, simulating a scenario in which news stories in an organization’s CMS are being discussed in a daily editorial meeting. They also interviewed 15 journalists to find out more about the dynamics behind the experiment’s results.

Journalists in the survey said they were skeptical that analytics should be used to guide story selection, but when faced with the experiment, they were significantly more likely to call for lower-performing stories to be pulled from the homepage. The effects of analytics were weaker, though, than topic, as journalists were more likely to advocate pulling consumer-oriented, “soft news” service journalism stories than politics or culture stories.

The interviews fleshed out some of these results. Journalists saw more of a role for analytics in post-publication actions such as demoting or removing stories than in choosing to cover or publish stories in the first place. Even then, they noted, “promoting brand-specific content or satisfying authors often takes precedence over strict adherence to audience analytics.” Overall, they concluded, journalists saw more of a need for balanced and diverse coverage during story selection, but used more analytical performance-based criteria to determine “de-selection.”

Journalists around the world are facing increasing levels of hostility, harassment, and even violence, both online and offline. These attacks on journalists have been sustained and concentrated enough to lead one scholar to characterize them as “mob censorship” — a form of “bottom-up, citizen vigilantism aimed at disciplining journalism.”

We also know that politicians have been relentless in their criticism of journalists, growing more aggressive in recent years. But there’s relatively little research to connect the two: Are politicians instigating this violence against journalists, and is this harassment and hostility catalyzing a chilling effect toward the media they’re attacking? Some previous research has found a connection, but this Flemish study is looking to pull much of this together — the pervasiveness, sources, and effects of violence — from journalists’ perspective.

The study used two rounds of surveys to gauge the opinions of Flemish political journalists (148 journalists in the first round, 39 in the second). The surveys found that even in a stable democracy with relatively high trust in the press, harassment was widespread and wasn’t limited to online spaces. Online harassment was unsurprisingly the most common form of hostility journalists experienced, with 83% reporting it, but 62% had experienced some sort of offline harassment as well, and more than a third had been threatened with criminal prosecution. Women were more likely to face sexual and verbal abuse, while men were more likely to receive physical violence, death threats, and threats of prosecution.

A notable amount of this harassment came directly from politicians and their staff: 55% of journalists reported receiving abuse from politicians or staff, though most said it happened rarely. Journalists said they were harassed by ordinary citizens and online trolls far more often, though they saw politicians, especially right-wing ones, as contributing to an undercurrent of aggression acted out by others.

Journalists said the hostility made them less likely to engage on social media and created ongoing stress and mental health concerns. The reports of self-censorship effects were much rarer, but the researchers still saw enough corrosive effects of harassment to maintain a concern about a “mob censorship” effect. As for the politicians’ role: “Politicians might not always be the direct aggressor,” they concluded, “but they often are the catalysts.”

One of the truisms of political campaigning is that everyone hates negative campaigning, but most candidates and parties do it anyway. In decades of research on negative campaigning, there’s no clear evidence that it’s effective in winning votes, and some limited evidence that it hurts people’s trust in government.

In this study, Amakoh was interested in a few wrinkles on these overall findings: The role of news coverage of negative campaigning, rather than the campaigning itself, and its effects on trust in legitimacy of election outcomes. Heavy media attention to negative campaigning, Amakoh hypothesized, “reinforces the perception that elections are driven by conflict rather than by rational decision-making,” leading to more skepticism of election results.

To test this, Amakoh used an 18-country African survey of trust in election outcomes, matching it with an international dataset of campaign coverage over the same period. He also examined the role of electoral management bodies in influencing election trust across those countries.

He found virtually across the board that coverage of negative campaigning was linked to lower trust in election outcomes, though this effect was substantially limited in countries with strong electoral management bodies. Perhaps surprisingly, the more media someone consumed, the less trust they were likely to show in election outcomes. (Much less surprisingly, supporting a winning candidate was linked to significantly higher levels of election trust.)

Ultimately, Amakoh concluded, while the news media has the ability to promote confidence in the electoral process by providing important information during a campaign, it also “has the power to erode the electorate’s trust in the electoral process through a focus on negative campaign messages.”