The accuracy and reproducibility of temperature measurements in solid materials under microwave heating are investigated in this work using two of the most celebrated temperature measurement techniques, namely fiber optic probes (FO) and infrared (IR) sensors. Two solid materials with a wide range of applications in heterogeneous catalysis and different microwave absorbing capabilities are examined: CeO2-ZrO2 and Al2O3 particles. We investigate a number of effects ranging from purely technical issues, such as the use of a glass probe guide, over process operation parameters, such as the kind and the volume of the heated sample, to measurement related issues, such as the exact location of the probe in the sample. In this frame, the FO and IR methods are benchmarked. It was found that when using bare FO probes, not only is their lifetime reduced but also the reproducibility of the results is compromised. Using a glass probe guide greatly assists in precise location of the probe in the sample resulting in more reproducible temperature measurements. The FO reproducibility, though, decreases with increasing temperature. Besides, contrary to conventional heating, the sample temperature decreases with decreasing sample mass (and volume) at constant irradiation power level, confirming the volumetric nature of microwave heating. Furthermore, a strongly non-uniform temperature field is developed in the reactor despite the use of a monomode cavity and small amounts of samples. These temperature variations depending on the volume and position can only by detected by FO. In contrast, IR, which actually measures temperature at the exterior of the reactor wall, remains nearly insensitive to them and consistently underestimates the real temperature in the reactor. The modeler and the experimentalist should be rather circumspect in accepting the IR output as a representative reactor temperature.