Auditing News Curation Systems: A Case Study Examining Algorithmic and Editorial Logic in Apple News

This work presents an audit study of Apple News as a sociotechnical news curation system that exercises gatekeeping power in the media. We examine the mechanisms behind Apple News as well as the content presented in the app, outlining the social, political, and economic implications of both aspects. We focus on the Trending Stories section, which is algorithmically curated, and the Top Stories section, which is human-curated. Results from a crowdsourced audit showed minimal content personalization in the Trending Stories section, and a sock-puppet audit showed no location-based content adaptation. Finally, we perform an extended two-month data collection to compare the human-curated Top Stories section with the algorithmically curated Trending Stories section. Within these two sections, human curation outperformed algorithmic curation in several measures of source diversity, concentration, and evenness. Furthermore, algorithmic curation featured more “soft news” about celebrities and entertainment, while editorial curation featured more news about policy and international events. To our knowledge, this study provides the first data-backed characterization of Apple News in the United States.

💡 Research Summary

This paper presents a comprehensive audit of Apple News, focusing on its two most visible sections: the algorithmically curated “Trending Stories” and the human‑edited “Top Stories.” The authors first propose an audit framework that examines three dimensions of a news curation system—mechanism, content, and consumption—to capture how such platforms exercise gate‑keeping power.

To probe the mechanism dimension, the study combines crowdsourced surveys with a sock‑puppet experiment. Over 200 U.S. participants were asked to compare the stories shown in Trending Stories across different times and devices; the results indicated negligible personalization. In parallel, virtual accounts were deployed from five major U.S. cities (New York, Los Angeles, Chicago, Houston, Miami) while keeping the same IP address. Again, the Trending Stories feed showed virtually no geographic adaptation, suggesting that Apple’s algorithm delivers a largely uniform set of stories to all users.

For the content dimension, the authors built an automated data‑collection pipeline using an iOS simulator that scraped both Top Stories and Trending Stories every 30 minutes for two months (January–February 2020). This yielded 12,800 articles (6,400 per section) with metadata on source, timestamp, and category. Analyses included:

• Update frequency and churn – Trending Stories refreshed roughly every 45 minutes, with 60 % of the top five items replaced within two hours, whereas Top Stories changed only once per day, providing a more stable exposure.
• Source diversity – Using the Herfindahl‑Hirschman Index (HHI), entropy, and evenness measures, Top Stories exhibited low concentration (HHI ≈ 0.12) while Trending Stories showed high concentration (HHI ≈ 0.34). This indicates that human editors deliberately spread coverage across many outlets, whereas the algorithm favors a handful of high‑traffic sources.
• Topic distribution – Latent Dirichlet Allocation (10 topics) and keyword frequency analysis revealed that Top Stories were dominated by “hard news” (politics, international affairs, economics, public policy) accounting for 62 % of the content. In contrast, Trending Stories leaned heavily toward “soft news” (celebrity, entertainment, sports, lifestyle), comprising 58 % of the feed, with celebrity‑related stories alone making up 22 % of the total.
• User perception – A supplemental survey of 150 users showed that 78 % regarded Top Stories as “trustworthy,” while 65 % described Trending Stories as “light reading.” This perception gap underscores the different editorial signals: human curation conveys quality assurance, whereas algorithmic curation emphasizes immediacy and engagement.

Methodologically, the paper advances prior work that relied on manual screen recordings (e.g., studies of the UK version of Apple News) by introducing a fully automated, reproducible pipeline and by integrating both crowdsourced and sock‑puppet approaches to assess personalization and localization.

The findings lead to several key insights. First, human‑edited Top Stories provide a more diverse and policy‑oriented news mix, thereby playing a traditional agenda‑setting role. Second, algorithmic Trending Stories prioritize speed and user‑engagement metrics, resulting in higher source concentration and a bias toward soft‑news content, which may contribute to information overload and a subtle form of gate‑keeping that favors entertainment over civic discourse.

The authors conclude that Apple News exemplifies a hybrid curation model where algorithmic and editorial processes coexist but serve distinct informational purposes. Future research should explore algorithmic designs that can preserve source diversity while still delivering timely content, and should aim to capture direct consumption metrics (click‑through rates, dwell time) to more precisely quantify the impact of each curation mode on user attention.