There is growing concern that online information searchers are overconfident and therefore largely search for information which reinforces their prior attitudes, blinded by confirmation bias. This study tests if this effect can be reduced in content aggregation platforms, when social tag clouds show popular topics among experts. We manipulated (1) confidence in prior attitudes, and (2) the credibility of the expert community that tagged the content. We found that both factors influence navigation in different ways. First, attitude confidence moderated the influence of prior attitudes when choosing how much attitude-consistent content in blog posts to read. When attitude confidence was high, prior attitudes were positively associated with selection of blog posts, when low, not positively associated. After navigation, when confidence was high, the content of attitude-consistent blog posts was more favourably evaluated, whereas when confidence was low, attitude inconsistent blog posts were more favourably evaluated. Second, source credibility moderated the influence of prior attitudes on tag selection. When source credibility was low, prior attitudes did guide tag selection, when high, they did not. With low source credibility, people selected more attitude-consistent content. The findings advance social tagging theories by showing that not only semantic associations, but also attitudes play a role when people select and process tags and related content. The findings also show that credibility and confidence have a different impact on different stages of information selection and evaluation. Whereas credibility is more important when switching among pages, attitude confidence is more important when reading and evaluating the content of one page.