Research report: Cues for you: a comparative test of agenda cueing and attitudes toward algorithmic news in the US and the UK

The study, published in the journal Communication and Change, tested whether the long-established agenda cueing effect survives when editorial recommendations are replaced by algorithmically personalized ones. The answer, at least in the American sample, was a clear split: journalistic cues raised the perceived importance of cued issues, while cues attributed to an algorithm did not produce uniform effects and instead split along ideological and political-interest lines.

Agenda cueing is a heuristic phenomenon in which signals about issue importance — a prominent headline placement, an editorial pick, a top slot in a news feed — shape which problems the public judges to matter most. As news distribution shifts from human-curated front pages to personalized recommendation systems, the study’s authors, Elizabeth Stoycheff and Robert B. Swetlic, set out to test the boundary conditions of this effect. Their experiment compared two countries with different regulatory postures toward artificial intelligence: the United States, with its commitment to accelerating competitive innovation, and the United Kingdom, which has adopted a more protective policy stance toward AI.

The design was a comparative online experiment. Participants in each country were randomly assigned to one of three conditions: a control group with no exposure, a condition in which agenda cues were attributed to journalistic selection, and a condition in which the same cues were presented as coming from an algorithmically personalized feed. The stimuli used the same five agenda items in both countries, deliberately chosen because they were salient in both the US and the UK at the time of data collection, in order to eliminate story topics as confounding variables. The lead story concerned the national budget.

The results in the United States supported the classic cueing hypothesis for journalistic cues. A one-way analysis of variance with planned contrasts found a significant difference between conditions in participants’ importance ratings of the lead budget issue, F(2, 387) = 2.96, p = .05. Respondents in the no-exposure control condition rated the budget as less important (M = 4.13, SD = 1.02) than those who saw the journalistic News Showcase condition (M = 4.41, SD = 0.77). For all cued issues taken together, the omnibus test was not significant, F(2, 387) = 2.17, p = .12, but planned contrasts showed that American participants exposed to the News Showcase condition rated these problems significantly higher than the control group, t(387) = 1.98, p < .05.

Crucially, the cueing effect was selective rather than general. A third analysis tested whether exposure to agenda cues simply raised concern about everything on the news menu. It did not: the treatment conditions did not significantly elevate the perceived importance of non-cued issues, F(2, 389) = 1.16, p = .32. This pattern matters theoretically, because it distinguishes genuine cueing from mere priming — the possibility that participants simply recalled whatever issues they had just seen in the experimental stimuli and rated them as important.

The UK results told a different story. Omnibus testing revealed no differences between conditions for lead story importance, F(2, 395) = 0.59, p = 0.56, cued issue importance, F(2, 388) = 0.71, p = 0.49, or non-cued issue importance, F(2, 390) = 0.69, p = 0.50. Unlike the Americans, British participants exposed to the journalistic News Showcase condition did not differ from the no-exposure control in their judgments of the budget issue, t(395) = 0.73, p = .46, or in overall cued issue importance, t(388) = −0.54, p = .59. The authors therefore concluded that only the US sample provided support for the journalistic cueing hypothesis.

The algorithmic condition, labeled Your Briefing, produced its own distinctive pattern. Comparisons between the Your Briefing condition and the control did not change cued problem importance judgments for US participants, t(387) = 0.63, p = .53, or UK participants, t(388) = −1.19, p = .24, on average. But moderation analyses revealed that the algorithmic cues were not uniformly inert. For liberal and politically interested Americans, Your Briefing cues were associated with increases in cued problem importance, t(385) = 2.05, p < .05. For Americans with low political interest, the algorithmically personalized feeds significantly lowered the perceived importance of cued problems, t(385) = −56.38, p < .001. In the US, partisanship and political interest appear to act as moderating forces for algorithmically personalized agenda cues — a finding the authors interpret as evidence that algorithmic curation interacts with user predispositions to fragment public agendas rather than unify them.

The study also delivered a direct test of the accessibility hypothesis, the idea that audiences merely recount the issues that are most readily available in memory after seeing them in the news. The researchers cued a non-problem — low teen pregnancies — alongside more acutely problematic issues in all treatment conditions. If participants were simply sampling accessible topics, teen pregnancy would have emerged as an issue of high concern in the treatment groups. It did not: participants in neither the US, F(2, 394) = 0.36, p = 0.70, nor the UK, F(2, 399) = 1.43, p = 0.24, perceived teen pregnancy as a greater problem after viewing news stimuli containing the issue. The authors read this as a refutation of the accessibility hypothesis and evidence that participants engaged in heuristic processing, using cues from journalists and AI systems to form their own judgments rather than mechanically repeating what they had seen. They suggest that heuristics such as personal relevance and ideological alignment in algorithmic feeds, and source credibility in journalistically prioritized feeds, help individuals decide which issues matter.

Attitudes toward artificial intelligence turned out not to be the driving force. Independent-samples t-tests found no significant differences between the US and UK samples in general attitudes toward AI, t(784) = 1.59, p = .11, or in attitudes toward algorithmic news specifically, t(791) = 1.63, p = .10, despite the two countries’ divergent regulatory approaches. Regression models predicting cued issue importance, fitted with a dummy-coded variable for the Your Briefing treatment and 5,000 bootstrap samples, included attitudes toward AI and algorithmic news attitudes as moderators, with political interest and ideology as controls. Neither general attitudes toward AI (US: β = −0.01, p = .91; UK: β = 0.07, p = .38) nor algorithmic news attitudes (US: β = −0.01, p = .89; UK: β = −0.05, p = .52) changed the cueing relationships. In the US, self-reported political interest and ideology filled that moderating role instead, which the authors attribute to the highly politicized nature of American news media, where partisan identity is strongly correlated with media trust and news credibility.

The authors are candid about the study’s limitations. Participants received a single message treatment, a conservative test of the cueing hypothesis. The UK news agenda was constructed to mirror the US stimuli rather than reflect issues endemic to Britain, and the authors suspect this design choice explains the null findings in the UK sample; the lead budget story was shown to Americans during the longest government shutdown in American history, while UK participants saw the same topic before a budget had even been announced. They urge caution against reading the UK nulls as anything other than a design flaw until further replications say otherwise, while noting that agenda cueing has yet to be empirically observed in a non-US sample. The personalization in the treatment was also rudimentary compared with the sophisticated, opaque systems built by large technology companies, and all participants saw the same five agenda items, underestimating the diversity of real personalized feeds. Even so, the authors argue the findings carry a clear implication: human editorial judgment still matters in setting the public’s priorities, and a hybrid model combining human oversight with algorithmic efficiency may help preserve the informational common ground on which democratic decision-making depends.

Subject of Research: Agenda cueing effects of journalistic versus algorithmic news curation in the US and UK

Article Title: Cues for you: a comparative test of agenda cueing and attitudes toward algorithmic news in the US and the UK

Article References: Stoycheff, E., & Swetlic, R. B. (2026). Cues for you: a comparative test of agenda cueing and attitudes toward algorithmic news in the US and the UK. Communication and Change, 2(1), Article 22. https://doi.org/10.1007/s44382-026-00042-3

Image Credits: AI Generated

DOI: 10.1007/s44382-026-00042-3

Keywords: agenda cueing, agenda setting, algorithmic news, news personalization, artificial intelligence, journalism, online experiment, political ideology, political interest, United States, United Kingdom, media effects