The local skincare market in Indonesia has experienced significant growth, accompanied by an increasing public interest in domestic products. The X social media platform is the leading platform for consumers to share experiences, opinions, and reviews of local skincare products. Consumer sentiment analysis is essential for understanding consumer preferences and perceptions of local brands, offering valuable insight that helps companies develop more targeted and effective marketing strategies for Indonesian consumers. This study conducts a sentiment analysis of local skincare brands using data from X in the Indonesian language. The IndoBERT model is applied as a feature extraction technique to capture the Indonesian language context more accurately, thus enhancing the sentiment classification process. This study evaluates the performance of Graph Neural Network (GNN) approaches for sentiment classification, focusing on the Graph Convolutional Network (GCN) and the Graph Attention Network (GAT). The experimental results showed that the combination of IndoBERT and GAT achieved a higher accuracy rate, 81%, compared to the combination of IndoBERT and GCN, which achieved an accuracy of 79%. These findings indicate that GAT is more effective in capturing the sentiment pattern in the Indonesian language data, outperforming the GCN model in sentiment classification within the specific context of local skincare review analysis.