Given the doubts about survey self-reports, media exposure research has turned to web tracking data. However, web tracking data is also biased. To improve the understanding of the quality of web tracking measures of media exposure, this paper estimates their validity and reliability. It additionally identifies design choices to optimize these. Using data from a cross-national three-wave survey, combined with web tracking, this paper conducts a multiverse analysis to assess the validity and reliability of +2,500 measures of media exposure. Results show an overall high reliability (0.86). In terms of predictive validity, the association between media exposure measures and political knowledge appears weak. This raises questions about the predictive validity of web tracking measures, and previous critiques to surveys self-reports. Additionally, results suggest that design choices impact the quality of web tracking measures. Methodologically, the paper presents a multiverse of measurements approach, improving the transparency of web tracking research.