AN ANALYSIS OF SIX ENGLISH TEXTS GENERATED BY ARTIFICIAL INTELLIGENCE (CHATGPT)

Abstract

This study aims to analyze the accuracy of generic structures and linguistic features of six English texts generated by ChatGPT, namely descriptive, narrative, recount, procedure, report, and explanation texts. This research employed a descriptive qualitative approach with the researcher as the main instrument. Data were collected through documentation by asking ChatGPT to generate one text for each of the six text types. The analysis was conducted by matching the generic structure and linguistic features of each text against standard genre theory. The findings indicate that the accuracy level of the generic structures in all six texts is very high. However, some errors were found in the linguistic features: the descriptive text contained inappropriate use of the Continuous Tense, and the procedure text contained the use of Future Tense in several sentences. Overall, the six texts generated by ChatGPT demonstrate excellent accuracy in generic structure but still require improvement in certain linguistic aspects. The limitation of this study is that it is confined to six text types and does not examine variations in prompts or different versions of ChatGPT. The implications of these findings can provide input for ChatGPT users, particularly students and educators, to be more critical in using AI-generated texts and to recognize the need for editing in linguistic aspects. The originality of this study lies in its contribution to evaluating the quality of academic texts generated by ChatGPT based on English genre conventions.