The objective of this study is to establish an ankylosing spondylitis (AS) thoracolumbar fracture finite element (FE) model and provide a proper posterior fixation choice from the biomechanical perspective. The ankylosing spondylitis T9-L5 FE model was built and the range of motion (ROM) was compared to previous studies. The L1 transverse fracture was simulated and was separately fixed by five different patterns. The pull force and yielding force of the screws, the von Mises stress of the internal fixation, and the displacement of fracture site were analyzed to evaluate the proper fixation pattern for thoracolumbar fracture of AS. ROM of AS model was obviously restricted comparing to the normal vertebral experimental data. All the fixation patterns can stabilize the fracture. At least four levels of fixation can reduce the von Mises stress of the internal fixation. Four levels fixation has a higher pull force than the six levels fixation. Skipped level fixation did not reduce the stress, pull force and yielding force. The kyphosis correction did not change the biomechanical load. At least 4 levels fixation was needed for AS thoracolumbar fracture. The cemented screws should be chosen in 4 levels fixation to increase the holding of the screws. The skipped fixation has no advantage. The kyphosis correction can be chosen after weighing the pros and cons.