Toxicity due to arsenic (As) in food has been mostly assessed by the determination of total As amount. Although the toxicity depends on the chemical forms, the sample can be determined to be safe if the total As concentration is less than the maximum allowable amounts of the toxic arsenic compounds. In this work, the accuracy of total As analysis for rice and oyster samples was significantly influenced by the matrix effect and the chemical forms. The matrix effect could be effectively corrected by employing the appropriate internal standard, tellurium. The As species such as As(V), As(III), monomethylarsonic acid, dimethylarsinic acid, and arsenobetaine showed significantly different sensitivities in the inductively coupled plasma mass spectrometry analysis for rice and oyster samples. This indicates that the total As determined without any information of the chemical species and matching sensitivities of samples and standards may be inaccurate.