News consumption is a central concept in political communication. Given the unreli- ability of survey self-reports, it is also notoriously difficult to measure. Recent advances in collecting the traces of people’s online activity have created potential to improve mea- surement of media habits in general. However, traces have been rarely used to test how people’s self-reported news exposure corresponds to actual news exposure. What is more, since traces are bound to a certain context in which they were collected—i.e. from a de- vice/browser that might or might not represent the person’s online and news consumption behavior—we must not take trace data as a panacea. In this pre-registered project relying on a multi-wave survey and several months of browsing data with a US sample (n = 2,219), we examine how well self-reported general news exposure, as well as self-reported outlet-specific exposures, corresponds to actual consumption. Beyond comparing the raw self-report to behavioral exposure, we also examine how correspondence changes when taking into account (a) how much people say they consume news online versus offline (b) how much they report to use other devices than the one from which we get traces (c) how much other people use the browser from which we get the traces. We find that generally, the correspondence between self-reports and traces is low—but can be improved some- what by taking into account the device context.