Determining Whether a Turkish Text is Produced by Artificial Intelligence or Human

Nowadays, with the acceleration of the flow of information over the internet, text content such as articles, blog posts and news are now produced in digital environments. Some of this content is produced by artificial intelligence rather than traditional authorship. This situation raises important questions about the reliability of text contents as well as the ease of access to information. The aim of this study is to determine whether an academic text was written by the ChatGPT artificial intelligence content creation tool or by a human. We received our data from the Dergipark platform, which contains Turkish academic articles. The data consists of summaries of Turkish articles. The dates of the articles taken from Dergipark are before the launch of Chatgpt (November 2022). In order to make a real distinction, we had to make sure that the data was actually written by a human. Distinguishing information such as the name and keywords of each article was given to Chatgpt 3.5 and the article summary was requested (tagged as ai). Abstracts of articles with the same distinctive information are also labeled human. The accuracy rate obtained in the study using LSTM is 0.8869 for the test set. In this way, it is aimed to make it easier to determine whether a written text was written by a human or an artificial intelligence text creator.