Photon counting x-ray imaging provides efficient rejection of the electronics noise, no pulse height (Swank) noise, less noise due to optimal photon energy weighting and the possibility of energy resolved image acquisition. These advantages apply also to CT when projection data are acquired using a photon counting detector. However, photon counting detectors assign a weighting factor of 1 to all detected photons whereas the weighting factor of a charge integrating detector is proportional to the energy of the detected photon. Therefore, data collected by photon counting and charge integrating detectors represent the 'hardening' of the photon beam passed through the object differently. This affects the beam hardening artefacts in the reconstructed CT images. This work represents the first comparative evaluation of the effect of photon counting, charge integrating and energy weighting photon detectors on beam hardening artefacts in CT. Beam hardening artefacts in CT images were evaluated for 20 cm and 14 cm diameter water cylinders with bone and low contrast inserts, at 120 kVp and 90 kVp x-ray tube voltages, respectively. It was shown that charge integrating results in 1.8% less beam hardening artefacts from bone inserts (i.e., CT numbers in the 'shadow' of the bone are less by 1.8% as compared to CT numbers over the periphery of the image), as compared to photon counting. However, optimal photon energy weighting, which provides highest SNR, results in 7.7% higher beam hardening artefacts from bone inserts as compared to photon counting. The magnitude of the 'cupping' artefacts was lower by 1% for charge integrating and higher by 6.1% for energy weighting acquisitions as compared to photon counting. Only the photon counting systems provide an accurate representation of the beam hardening effect due to its flat energy weighting. Because of their energy dependent weighting factors, the charge integrating and energy weighting systems do not provide accurate representation of the beam hardening effect.