Before the 2013 German federal election, 121 participants completed a 2-session online study (which paralleled a U.S. study before the 2012 presidential election). They browsed online search results pertaining to 4 political issues while selective exposure was unobtrusively measured. In a 4×2×2 (topic×issue stance×source credibility) within-subjects design, the search results indicated either issue support or opposition, associated with low- or high-credibility sources. Hypotheses were derived from cognitive dissonance, approach-avoidance, and motivated cognition models. Findings yielded a confirmation bias. Attitude-consistent exposure uniformly reinforced attitudes; attitude-discrepant exposure uniformly weakened attitudes. Analyses with parallel U.S. data showed a stronger confirmation bias in the United States than in Germany.