Fracture zone evolution is investigated using dynamic models that allow the fault zones to freely slip. This is an improvement over past formulations where bathymetric offsets were imposed kinematically. The models use a viscoelastoplastic rheology that incorporates the influence of fault friction on fracture zone slip history. Using viscoelastic plates, we assess the role of small-scale convection on removal of the lowermost thermal lithosphere beneath fracture zones. Through a comparison of synthetic gravity to free-air gravity across fracture zones we find that the amplitude of the gravity jump across fracture zones is best fit by models with weak faults that have depth-averaged yield strengths <10 MPa. Fracture zones with such low strengths can convert to subduction zones with ~100 km of convergence. Many fracture zones do not fit plate subsidence models with locked or slipping faults but are better fit by systems that are tectonically deformed by modest amounts of extension.