Volcanic rocks are considered reliable recorders of past changes in the Earth's magnetic field. Recent flows, however, sometimes fail to produce the known magnetic field at the time of cooling. Previous research on Mt. Etna suggests paleomagnetic data might not be accurately recorded. Here we test the accuracy of paleomagnetic data obtained from Mt. Etna lavas by comparing paleomagnetic data from historical flows to direct measurements of the magnetic field above the current topography. The inclinations and intensities in both data sets are biased toward lower values, while there is no such trend for the declination. Inclinations are on average 2.9° lower than expected; intensities are on average 8.8 µT lower. The deviations from the expected values depend on the height above the flow. Moreover, the inclinations and intensities vary as a function of topography. Both are higher above ridges and lower in gullies; the variations within a site are up to 14.1° in inclination and 12.9 µT for intensity. To suppress this paleomagnetic data bias it is important to take samples several meters apart and from different parts of the flow whenever possible. While this leads to a higher degree of scatter in paleodirections, the results better represent the Earth's magnetic field at the time of cooling. This emphasizes the importance of reporting paleomagnetic sampling strategies in detail. Plain Language Summary Paleomagnetic data from lavas is routinely used in the Earth Sciences to for example, reconstruct the past behavior of the Earth's magnetic field, or make models of past plate motions. Very young flows for which the ambient magnetic field at the time of cooling is known, however, sometimes fail to produce the known reference values. Lava flows from Mt. Etna are extensively studied in the past and the paleomagnetic data obtained does often not agree with the known magnetic field value in which the lavas cooled. Here we show that the topography of volcanic terrain may influence the magnetic signal of new, overlying, flows, and we make recommendations for sampling strategies that suppress these terrain effects as much as possible. Key Points Paleomagnetic data from Mt. Etna does often not reproduce the known geomagnetic field well Local magnetic anomalies explain bias in paleomagnetic data as function of topography Optimizing the paleomagnetic sampling strategy may suppress this bias in paleomagnetic data