AI Browsers (Chrome, Edge, Perplexity) Cut Political Bias in News Summaries
New study of 41K summaries shows AI browsers neutralize negativity and partisan slant
A large-scale audit by Yan Xia and colleagues evaluated how AI-powered browsers reshape news before millions read it. Using 13,777 articles from 15 U.S. news outlets, they generated 41,331 summaries across Google Chrome (Gemini), Microsoft Edge (Copilot), and Perplexity Comet. The study found that all three browsers produce broadly accurate summaries. However, they systematically alter the content: reducing ideological bias, partisan stances, negativity, anger, and fear, while improving clarity and reducing personal tone. These patterns persist across browsers, outlet ideologies, and topics, with only minor variations. The authors describe AI browsers as 'editorial intermediaries' that transform news in consistent, measurable ways.
This systematic reshaping has profound implications for democratic discourse and AI governance. By dampening negativity and bias, AI summaries could reduce political polarization and negative affect among readers. Yet the same mechanism raises concerns about homogenized perspectives and the loss of journalistic nuance. As browser-based AI reaches millions of users, understanding these editorial effects becomes critical for policymakers, news organizations, and the public. The study provides the first empirical evidence of how AI browsers alter news at scale, calling for transparency and oversight of these increasingly influential systems.
- Analyzed 13,777 articles from 15 U.S. news outlets, generating 41,331 summaries across 3 browsers (Chrome Gemini, Edge Copilot, Perplexity Comet)
- AI summaries are broadly accurate but consistently reduce ideological bias, partisan stances, negativity, anger, and fear
- Patterns hold across different browsers, outlet ideologies, and topics, positioning AI browsers as systematic editorial intermediaries
Why It Matters
AI browsers now act as invisible news gatekeepers for millions, potentially reducing polarization but also filtering journalistic nuance.