Eye tracking is becoming increasingly popular in physics education research (PER). As technology has advanced considerably in recent years and has become more user friendly, it is anticipated that eye tracking will play an increasingly significant role in assessing student learning at the process level in future studies. The main objective of this systematic review is to summarize the current status quo regarding eye tracking in PER and reviewing (a) the dissemination, (b) the methodological implementation, and (c) the insights provided by eye tracking in PER. We identified 33 journal articles, published between March 2005 and April 2021, that used eye tracking for original empirical research in the area of physics education. The results reveal that although eye tracking has been used in many different areas of physics, a clear focus on mechanics is evident, particularly for measuring visual attention in assessment scenarios like problem solving. While a high methodological rigor in the selection and analysis of the visual stimuli was apparent, only a few studies have provided a complete documentation of the technological implementation (e.g., movement restrictions, accuracy, and calibration information) and a theoretical embedding for interpreting eye-tracking data. To synthesize the results of the different studies, we created an inductive category system in accordance with the considered independent variables of the studies. Accordingly, visual attention was most frequently compared between levels of performance (correct vs incorrect or high vs low achievers), thereby leading to performance-discriminating factors of eye movement across studies. Furthermore, learners' eye movements were compared across different stimuli, different time points, or between student groups to inform multimedia design and shed light on students' learning progression. In summary, eye tracking is particularly useful for studying processes in different domains that are relevant to PER.