AI news summaries leave Australian media behind on Copilot

 JournalismPakistan.com |  Published: 25 January 2026 |  JP Global Monitoring

University of Sydney researchers analyzed Copilot's news replies and found only about one‑fifth linked to Australian outlets; citations favored major national and international organizations, while regional and independent newsrooms were largely sidelined.

Summary

SYDNEY, Australia — A new study by researchers at the University of Sydney has found that Microsoft Copilot’s AI‑generated news summaries tend to under‑represent Australian local and independent media outlets, instead favoring large United States and European news organizations.

The research, led by Dr. Timothy Koskie of the Centre for AI, Trust and Governance, analysed hundreds of Copilot news responses and found that only about one‑fifth included links to Australian media sources. Those that did cite domestic outlets were mainly major national publishers, such as Nine and the ABC, with regional and independent outlets largely absent.

Copilot’s summaries also showed a pattern of disproportionately citing international outlets such as CNN and the BBC, even when queries originated in an Australian context, researchers said. Local journalists’ bylines were often omitted entirely, with AI outputs providing summaries that lacked explicit sourcing or credit for reporting across smaller outlets.

Impact on local journalism

Experts behind the study warn that this pattern could deepen structural challenges already facing Australian media, where independent and regional newsrooms struggle with limited resources and audience reach. By privileging large global outlets in AI news outputs, Copilot may reduce referral traffic to smaller publishers’ websites, erode byline visibility, and weaken economic incentives for local reporting.

The researchers also highlighted the risk that such output dynamics may contribute to news deserts outside major urban centres, where local coverage is already scarce. Without adequate representation in algorithm‑driven news tools, regional communities may gain less exposure to information that affects their daily lives, from municipal governance decisions to emergency alerts.

Calls for policy and design changes

The Sydney team has urged policymakers and AI developers to consider interventions that encourage greater geographic sensitivity in AI news tools. Suggestions include adapting news media bargaining codes and requiring platforms to account for local publishing diversity when generating summaries. They argue that such measures could help balance global reach with the need to sustain local news ecosystems that underpin democratic participation and civic engagement.

Limitations and broader context

Microsoft acknowledges that its Copilot models may reflect biases embedded in underlying training data and that outputs can occasionally over‑ or under‑represent particular groups or perspectives. Company documentation notes ongoing efforts to refine safety systems and counteract unwanted skew in AI responses, though it also emphasises the probabilistic nature of large‑language‑model outputs.

WHY THIS MATTERS: This study highlights a growing concern among journalists and media professionals about how AI tools shape news discovery and consumption. For Pakistani newsrooms grappling with their own challenges around media diversity, local language coverage, and digital audience engagement, the findings underscore the importance of advocating for transparency and equitable representation in AI‑driven news systems that could shape public understanding and influence revenue ecosystems for smaller publishers.

ATTRIBUTION: Reporting is based on analysis by the University of Sydney and related coverage in The Guardian and News Minimalist.

PHOTO: By Tawanda Razika from Pixabay