Abstract— The rapidly growing social media platform X as one of the most popular platforms presents new challenges for its users. The lack of supervision and control over the content shared on this platform raises serious problems, one of which is related to the credibility of information. The credibility of information on social media is a crucial aspect that needs to be considered to assess the level of user trust in information. Therefore, this study aims to develop a system that is able to detect the credibility of information on content shared, especially on the social media platform X. This system is designed by implementing a deep learning approach using the Gated Recurrent Unit (GRU) algorithm as a classification model, adding feature extraction using TF-IDF and feature expansion using GloVe, and using Artificial Bee Colony (ABC) as an optimizer. The application of GloVe in this study aims to expand features by finding similarities in words to enrich the k features, while ABC optimization aims to find the best parameters to improve the performance of the GRU classification. The combination of all these methods produces a final accuracy of 77.80%, which shows that feature expansion using GloVe and optimization with ABC can significantly improve model performance. This study effectively creates a reliable method and shows positive results in assessing the credibility of information through the mentioned techniques.