Southeast Asia’s cultural heritage is deeply embedded in its traditional textiles, such as songket, limar, batik, and ikat, each reflecting centuries of artistic evolution and cross-cultural interactions. However, preserving and analysing these intricate motifs remains a challenge. This study leverages deep learning and unsupervised machine learning to systematically classify and trace the phylogeny of traditional ASEAN textile patterns. Our methodology involves an autoencoder-based feature extraction followed by K-Means clustering to categorise textile motifs based on their latent representations. A dataset of traditional ASEAN textiles was collected and preprocessed by resizing images to 128 × 128 pixels and normalising pixel values. A convolutional autoencoder, built using TensorFlow’s Keras API, was trained to encode textile images into a low-dimensional latent space. The encoder comprised convolutional layers with ReLU activation and max-pooling, while the decoder reconstructed the original patterns. Once trained, the encoder extracted latent features, which were then clustered using K-Means with cluster numbers ranging from 5 to 15 to identify optimal groupings. Representative images from each cluster were visualised to verify the homogeneity and distinctiveness of motifs. By systematically clustering traditional textile patterns, this study provides a computational framework for motif classification, heritage conservation, and cross-cultural analysis. The findings contribute to digital archiving, authenticity verification, and motif evolution studies, supporting efforts in intangible heritage preservation and computational anthropology. This work showcases the power of AI in cultural studies and offers scalable tools for textile heritage analysis in the digital age. Further research aims to increase the dimensions of the artefacts from 2D textiles to 3D keris (dagger) and 4D tanjak (headdress).