Although vegetation restoration controls gully head retreat (GHR), gully heads (GHs) could still retreat on vegetation-covered lands under intensifying rainstorms induced by global climate change. However, the GHR processes and mechanisms for GHs on vegetation-covered lands are unclear. A series of simulated experiments were conducted on a typical grassland (Medicago sativa L. dominated) under different inflow rates (3.6–9.6 m3 h−1) to explore the hydraulic and mass movement properties, morphology evolution, and sediment yield process of GHs induced by surface flow. The results show that the jet flow dominated and accounted for 51.94%–63.33% of the inflow volume; the on-wall flow (including seepage flow) rate linearly increased from 1.40 to 3.08 m3 h−1, while its proportion decreased with increasing inflow rate. A scour hole could gradually develop on the headwall due to the on-wall flow scouring, seepage, and scour hole collapse, resulting in an overhanging mass. A critical scour hole size was observed when the gully headwall overhanging mass collapse (GHOC) occurred under different inflow rates. Additionally, GHOC occurred earlier under the higher inflow rate, which was associated with the more significant increasing rate of the scour hole size under the higher on-wall flow rate. The plunge pool obtained the maximum depth and length before the GHOC, and the higher the inflow rate that was applied, the higher the maximum depth and length that were observed. The sediment yield rate decreased over time with fluctuations, and the maximum values resulting from the GHOCs were 1.60–5.06 times the initial values. Plunge pool erosion had limited effects on GHR, while scour hole development on the headwall and subsequent overhanging mass collapse played critical roles in GHR on grassland under surface flow.