Many have argued that the algorithmic personalization of information provision can lead to segregation and political polarization. We test this conjecture for the case of online news search via Google. Our study design rests on 300 US-located synthetic web users ("bots") with human-like appearance and randomized individual characteristics who were active from June to December 2021. We find that web search personalization increases informational segregation in news-related search results between users with liberal vs conservative political preferences, as well as between users with and without privacy settings. However, our evidence suggests that an increase in polarization due to personalized search results is rather unlikely. On the contrary, due to the overall dominance of a few rather centric news sites in news-related search results, ideology-updating users tend to move towards the political center.
