Forest roads play a crucial role in commercial forestry. Most of them are unpaved and exposed to increased surface runoff and the occurrence of soil erosion. Therefore, these roads constantly require repairs and maintenance, which are associated with the high cost of earthworks. Soil water repellency causes retardation of water infiltration into soil, which affects the generation of surface runoff, leading to the destruction of the active road layer. This phenomenon is well recognized in forest habitat soils, but poorly investigated in terms of soils constituting roads and communication routes. Therefore, the aims of this research were to determine the severity and persistence of potential water repellency in the case of soil collected from two unpaved roads in coniferous forest and to investigate the dependence between the severity of soil water repellency and its persistence. Soil samples (Dystric Brunic Arenosols) were collected from three points of the roadway, i.e., centre line, rut and roadside, from two different sites located in fresh coniferous and fresh mixed coniferous forest (Poland). The severity of soil water repellency was assessed, based on measured values of the wetting contact angle (Wilhelmy plate method) and the critical and 90° surface tensions. For this purpose, contact angle measurements of the solid phase, wetted by water–ethanol mixtures, were used and data results were analysed using the Zisman approach. The persistence of potential soil water repellency was assessed using the water drop penetration time test. The severity and persistence of soil water repellency depend on the forest habitat type surrounding the road and point locations on the road cross section. The soil samples collected from the roadside were more water repellent than samples taken from the centre line or rut. A strong relationship was found between the water drop penetration time and the advancing contact angle, the critical surface tension and the 90° surface tension.