The thixotropy of lime-modified loess is a key engineering problem in large-scale mountain levelling and urban construction on the Loess Plateau in China. We analyse the thixotropic factors and establish thixotropic models of modified loess at macroscales and microscales to interpret the evolution of the thixotropic mechanism of lime-modified loess. A custom-made volume-preserving thixotropic instrument is used to eliminate the influence of consolidation deformation on thixotropy and simulate soil consolidation in the field. Consolidated undrained triaxial tests, nuclear magnetic resonance analysis, and electron microscopy are used to investigate the thixotropy of soils with different thixotropic periods (0 days, 7 days, 14 days, 21 days, 42 days, 84 days, 126 days, and 168 days). The results show that the failure strength increases and the growth rate decreases with the thixotropic period length. The failure strength increases rapidly in the early thixotropic stage; the inflexion point occurs at 21 days, and stabilisation is observed at about 42 days. The internal friction angle and cohesive force increase over time, the cohesive force increased more obviously, which was 2.94 times of the initial thixotropic period, the increase in internal friction angle is within 4 degrees. The pore distribution is more uniform at the microscopic level, and large and small pores are transformed into medium pores over time. As the thixotropic period increases, the amount of cementitious material generated in the modified loess and the cementation degree increase, and the number of surface pits and large pores on the particles substantially decreases, resulting in numerous flower-shaped and grid structures. The thixotropic mechanism of modified loess consists of pore homogenisation, gravitational repulsion between particles, and cementation caused by the lime reaction.