Different computer simulated persons (CSPs) have been developed to represent occupants for indoor airflow simulation using computational fluid dynamics (CFD). Simple CSPs are preferred in order to avoid very fine grids and long computing time. However, how CSPs with different complexity affect the accuracy of CFD simulation is not well studied yet. This paper intends to investigate quantitatively the simulation accuracy due to CSP simplification. A detailed, human-like CSP and three different simplification strategies including an overly simplified rectangular box CSP, a simple CSP using several cuboids to represent different parts of a person, and an improved one in which the total heat of different parts (head, body, arms, legs) matches that of the detailed CSP are proposed. The simulation results using simplified CSPs are compared with the benchmark experimental data and simulated results using the detailed CSP in a displacement ventilation case. Results show that all the CSPs, simple or complex, yield acceptable simulation results for the global field study. For the micro-environment around the person, there seems no significant benefit by adopting very complex CSP geometry either. Instead, attention should be paid to ensure the heat and pollutant source consistence of the simplified CSP over the actual human (detailed CSP).Units: SI