We examine the introduction of automatic air pollution monitoring, which is a central feature of China’s “war on pollution.” Exploiting 654 regression discontinuity designs based on city-level variation in the day that monitoring was automated, we find that reported PM10 concentrations increased by 35% immediately post–automation and were sustained. City-level variation in underreporting is negatively correlated with income per capita and positively correlated with true pre-automation PM10 concentrations. Further, automation’s introduction increased online searches for face masks and air filters, suggesting that the biased and imperfect pre-automation information imposed welfare costs by leading to suboptimal purchases of protective goods.