The experimental rotational constants are generally determined with a very high precision. However, before using them for a structure determination, they have to be submitted to several corrections (centrifugal, rovibrational, electronic) which considerably reduce their accuracy. The rovibrational correction is by far the largest one. It can be estimated from a cubic force field usually computed ab initio. Its accuracy obviously depends on the molecule and on the level of theory used for the computation of the force field. However, it appears that, in most cases, the MP2 method with a basis set of triple zeta quality give results which are accurate enough for a structure determination. In this case the accuracy of the equilibrium rotational constants is a few percent (from 2% up to about 10%) of the rovibrational correction. It also appears that the accuracy of the A e rotational constant is often the worst one.