GEANT4 is a Monte Carlo code originally implemented for high-energy physics applications and is well known for particle transport at high energies. The capacity of GEANT4 to simulate neutron transport in the thermal energy region is not equally well known. The aim of this article is to compare MCNP, a code commonly used in low energy neutron transport calculations and GEANT4 with experimental results and select the suitable code for gadolinium neutron capture applications. To account for the thermal neutron scattering from chemically bound atoms [S(alpha,beta)] in biological materials a comparison of thermal neutron fluence in tissue-like poly(methylmethacrylate) phantom is made with MCNP4B, GEANT4 6.0 patch1, and measurements from the neutron capture therapy (NCT) facility at the Studsvik, Sweden. The fluence measurements agreed with MCNP calculated results considering S(alpha,beta). The location of the thermal neutron peak calculated with MCNP without S(alpha,beta) and GEANT4 is shifted by about 0.5 cm towards a shallower depth and is 25%-30% lower in amplitude. Dose distribution from the gadolinium neutron capture reaction is then simulated by MCNP and compared with measured data. The simulations made by MCNP agree well with experimental results. As long as thermal neutron scattering from chemically bound atoms are not included in GEANT4 it is not suitable for NCT applications.