Personalization and Outrage: Examining Sentiment Bias in Google Discover’s US Political Coverage

Abstract

Google Discover is a primary source of traffic for many news publishers across the globe. The United States, in this context, is no exception. In the age of advanced internet systems, driven by AI and algorithms, users’ digital behavior determines what content is promoted to them by platforms such as Google. This, in turn, determines how people perceive social and political realities. But, while positioned as a personalized and interest-based news recommendation system, Google Discover’s content curation is neither neutral nor free from bias. Instead, it reconfigures existing biases through algorithmic personalization, reinforcing the argument that algorithms are socially embedded and value-laden. Building on prior studies that have found predominantly positive sentiment in Google News, this study shifts focus to Google Discover to understand if similar patterns persist. Using sentiment analysis conducted via Google Cloud’s analyze Sentiment method, this research analyses the emotional tone of political news headlines promoted on Google Discover in the United States. Sentiment scores classify headline tone as positive, negative, or neutral, thereby framing media content as algorithmically curated rather than as audience reaction. By employing qualitative content analysis, this study examines how platforms privilege political news headlines with predominantly negative sentiment. It also aims to explore how algorithmically amplified news stories on Google Discover promote a specific group of people and their political ideologies. The study is grounded in Networked Agenda-Setting theory and the Platform Power concept to understand the underlying implications of giving space
to certain news on Google Discover over others and showcase how platforms influence audience choices. The study raises concerns about balanced news consumption in highly personalised news environments and points toward the need for greater transparency. Future research could extend this analysis through cross-platform or cross-national comparisons to further explore sentiment bias in algorithmic news ecosystems.