Google News and machine gatekeepers: algorithmic personalisation, filter bubbles and homogeneity in online news search


Machines are increasingly aiding or replacing humans in journalistic work, particularly in news distribution. The algorithms underpinning search results and recommendations - the cornerstone of navigating the web – raise fundamental normative questions over the role of machines as news gatekeepers (Napoli, 2015; Nechushtai and Lewis 2019). Here, existent research into personalisation has found conflicting results, with some finding evidence of personalisation (Hannak et al. 2013), and others finding that search engines expose people to news that is opposite of their political opinion (Flaxman et al. 2016).

Through a mixed methods research design, in this paper we address normative aspects of news recommendation engines by examining whether search personalisation, diversity and filter bubbles are evident on Google News in the UK. Firstly, in a quasi-experimental design borrowing from Nechushtai and Lewis (2019), we asked a diverse set of participants (N =86), to search Google News (though their personal Google accounts) for four search terms based on the two main party leaders, and two contested political topics and report the first five stories they were recommended on each term in our survey.

We found that personalisation was evident based on geo-location, but did not find a correlation to any other variable, including participants’ previous online news-searching behaviour and political leaning; therefore challenging the claim that news search algorithms result in echo chambers. Further, we found a high degree of homogeneity in news search results. The top nine news sources recommended by Google made up 75% of the 775 recommendations and are all either print-based or broadcast legacy media. New digital-only or alternative news sources barely figured, suggesting that Google News algorithms do little to disrupt existing industry power dynamics.

Secondly – and to further examine the diversity of search results – we conducted a manual content analysis on the stories recommended by Google News for our search terms (N=775), focusing on their favourability towards the search term in question (i.e. the politician or political issue). Results showed that while there was little relationship between the slant of the article and the political leanings of participants, there was one exception, where self-identified right-wing participants were more likely to see unfavourable stories about immigration. This reopens the question of filter bubbles for certain news consumers. Findings are discussed in relation to ongoing debates around algorithmic news cultures, the (ir)relevance of journalism and public knowledge of political topics.