The determination of near‐surface (vadose zone and slightly below) fault locations and geometries is important because assessment of ground rupture, strong shaking, geologic slip rates, and rupture histories occurs at shallow depths. However, seismic imaging of fault zones at shallow depths can be difficult due to near‐surface complexities, such as weathering, groundwater saturation, massive (nonlayered) rocks, and vertically layered strata. Combined P‐ and S‐wave seismic‐refraction tomography data can overcome many of the near‐surface, fault‐zone seismic‐imaging problems because of differences in the responses of elastic (bulk and shear) moduli of P and S waves to shallow‐depth, fault‐zone properties. We show that high‐resolution refraction tomography images of P‐ to S‐wave velocity ratios (V_P/V_S) can reliably identify near‐surface faults. We demonstrate this method using tomography images of the San Andreas fault (SAF) surface‐rupture zone associated with the 18 April 1906 ∼M 7.9 San Francisco earthquake on the San Francisco peninsula in California. There, the SAF cuts through Franciscan mélange, which consists of an incoherent assemblage of greywacke, chert, greenstone, and serpentinite. A near‐vertical zone (∼75° northeast dip) of high P‐wave velocities (up to 3000  m/s), low S‐wave velocities (∼150–600  m/s), high V_P/V_S ratios (4–8.8), and high Poisson’s ratios (0.44–0.49) characterizes the main surface‐rupture zone to a depth of about 20 m and is consistent with nearby trench observations. We suggest that the combined V_P/V_Simaging approach can reliably identify most near‐surface fault zones in locations where many other seismic methods cannot be applied.