It is a difficult and hot issue in the hydrological studies of arid areas to choose suitable methods to evaluate the recharge of atmospheric precipitation to groundwater and its response to climate change in desert areas. This study reviews the theories and problems of vadose (unsaturated)-zone tracing methods selected by predecessors in hydrological studies and takes the deserts in middle latitudes of northern China as an example to extract decadal, centennial, and millennian information of atmospheric precipitation to groundwater recharge on a regional scale since the late Holocene. The fluctuations of atmospheric precipitation and chronological sequences of desert unsaturated zone were estimated by using the chlorine mass balance (CMB) theory. It indicates that the Badain Jaran Desert in the central Alashan Plateau and the surrounding Gobi deserts have experienced fluctuations of groundwater recharge on a centennial scale during the late Holocene period from about 700 to 2000 years ago. Multiple CMB profile records can identify four periods of relative wetness (1330–1430, 1500–1620, 1700–1780, and 1950–1990) and three periods of relative drought (1430–1500, 1620–1700, and 1900–1950) over the past millennium. These records are consistent with other paleoclimatic records in the northern margin of the Qinghai-Tibet Plateau, and relatively correspond to those in the eastern part of China. This indicates that groundwater recharge in the Alashan Plateau broadly reflects the degree of climatic variability in northwest China over the centennial scale and may be affected by the changes in the intensity of the East Asian summer monsoon. The estimated average recharge rate of precipitation in the Alashan Plateau in the last millennium is about 1.3~2.6 mm/a, which brings new geological evidence for understanding the source of groundwater recharge in the region but is quite different from other environmental records. It should be noted that there are uncertainties in the CMB records of the vadose zone profiles, mainly due to the assumption of atmospheric Cl input in the CMB estimation and the selection of the homogeneous vadose profile (piston flow). This study suggests that this uncertainty and its error should be extensively tested in the future by comparing deterministic data (such as regional reference stations) with large-scale random atmospheric Cl input backgrounds.