Do wildfires affect the educational outcomes of elementary and middle school students? Are these effects immediate and transitory, or do they persist over time? As wildfire seasons across the globe grow longer and more severe, these questions are becoming increasingly important. Here, we use student testing data from Colorado to investigate immediate and persistent causal impacts of the destructive 2020 wildfire season on elementary and middle school test participation and competency rates from 2021 to 2023. School disruptions are proxied using near real-time satellite observations of smoke plumes on school days. We demonstrate that recovery of educational outcomes to pre-pandemic levels is hindered by disasters, with affected schools reporting test participation rates about 2-3 percentage points below those of their peers even two years after the disaster. We also show that students at all grade levels are affected immediately by wildfire-induced school disruptions, but effects are most persistent for younger students in urban settings.