This paper reports comparative experimental evidence on whether and how citizens hold democratic governments accountable for their performance during the Covid-19 pandemic. Drawing on divergent theoretical perspectives, our pre-registered survey experiment conducted in three countries is designed to, first, test whether exogenously varying cross-national benchmarking information affects respondents’ overall evaluation of how their country’s government has handled the virus. Second, it tests whether pre-treatment satisfaction with the government affect the choice the benchmark. The experiments were embedded in nationally representative internet surveys in France, Germany and the UK. We find that exogenously varying benchmarking information has consistent effects on evaluations, consistent with standard theories of accountability. However, when being able to choose their information, respondents strongly select a benchmark that is in line with their prior view of the government. This is in line with theories emphasizing behavioral constraints. Taken together, these results suggest that endogenous information acquisition limits political accountability during the pandemic. This effect is likely to be most relevant when political polarization is high.