Significance Education is increasingly delivered online, but are students actually paying attention? Here we demonstrate that efficacy of video instruction can be assessed remotely with standard web cameras. Specifically, we show that attentive students have similar eye movements when watching instructional videos and that synchronization of eye movements is a good predictor of individual learning performance. Measuring synchronization of eye movements while preserving privacy, as we have shown here, has the potential to make online education adaptive to attentional state and advance mechanistic studies on the efficacy of different online education formats. Attention has become a commodity online. With the increasing abundance of video content, remote sensing of attention at scale may be relevant beyond education, including entertainment, advertising, and politics.