In this paper we reanalyze the centennial trend estimates of a number of selected geomagnetic indices. We show that the peculiar latitudinal ordering of century long trends is mainly an artifact related to different scales. Using a scaling method that takes into account the statistical dispersion of the variables alleviates most of the discrepancies associated with latitude dependence. We demonstrate that geomagnetic activity had a roughly homogenous increase over the Northern Hemisphere during 1901–2000 of about 23%, as registered by digital measures. On the other hand, midlatitude data are shown to possess a distinct disproportion dominantly around solar cycle 15. We also demonstrate that analog indices tend to consistently depict lower activity levels in the early decades, thus considerably larger overall increases of about 32% during the twentieth century. This is likely due to the typically conservative approach in quantifying irregular activity (the K value) in those early years with a large number of quiet days. The recently corrected aa index shows an increase on par with similarly analog but local measures at Niemegk and Sodankylä stations. These results show that after its vital calibration, the widely used aa index does quantify reasonably well the global centennial increase in geomagnetic activity. Key Points Long‐term trends registered at latitudes are similar Midlatitude data possess disproportion around cycle 15 Analog magnetic indices tend to depict larger trends