In the present study, two sets of scripts from primary school students were collected, one written in English and the other in their native Chinese on the same topic. The Chinese scripts were translated into English by Google Translate (GT) and compared with the scripts written in English. Sentences in the two sets of passages that were clearly parallel in meaning were then extracted and compared for accuracy, vocabulary, substance, and length. Findings revealed that in some cases the GT versions (i.e., those originally written in the native tongue) displayed language that was significantly better that what the students produced when writing in English. Patterns of improvement are analyzed and discussed.