The carbonate clumped isotope (∆47) thermometer is a powerful tool for temperature reconstruction based on the thermodynamical 13C—18O bonding preference. Accurate interpretation of temperatures requires preservation of original carbonate ∆47 signatures, which are susceptible to resetting at elevated temperatures during burial. Kinetic models of carbonate 13C—18O bond reordering derived from experimental studies constrain the evaluation of ∆47 preservation and the reconstruction of thermal histories and initial temperatures for reordered carbonates. However, application of these models to different carbonates is limited because kinetic parameters vary with materials and are believed to be affected by factors such as crystal structure, cation, trace element composition, crystal defects, and internal water whose roles in reordering are still not well-understood. Here we present reordering data for strontianite and witherite from heating experiments and evaluate the effects of crystal structure, cation chemistry, cation vacancy and substitution, and internal water on reordering using constrained ab-initio molecular dynamics (cAIMD) simulations. We show that strontianite and witherite reorder significantly faster than other carbonates. Comparing experimental studies yields the following reordering rate sequence: strontianite > witherite > aragonite > calcite > dolomite. The general trend in Arrhenius activation energy for reordering is strontianite < witherite < aragonite < dolomite < calcite (with calcite values based on the average of individual materials). Our cAIMD simulations reveal a similar trend in free energy of activation for oxygen exchange in these carbonates. We suggest that orthorhombic carbonates are more susceptible to reordering than trigonal carbonates because of the shorter distance between carbonate groups and because the weaker M—O interaction in orthorhombic carbonates reduces the structural distortion and the energy cost during transition state formation. For carbonates sharing the same structure, the weak M—O bonding can be compensated by a large unit cell volume, leading to a non-systematic relationship between the free energy of activation and the cation radius. In simulations with lattice-bound water, the free energy of activation is greatly reduced compared with cation vacancy and substitution. The influence of fluid inclusions on measured reordering rates depends on the fraction of carbonates that can exchange with water as suggested by strontianite δ¹⁸O data. Our results serve as a starting point for a more sophisticated but flexible kinetic model containing material-specific factors (e.g., substituted ions, structure water, etc.) for reordering in abiotic and biotic carbonates.