Abstract
With the rise of user-generated content, ensuring the authenticity and originality of online information has become increasingly challenging. Artificial intelligence (AI) and Natural Language Processing (NLP) play a crucial role in large-scale content analysis and moderation. However, the widespread use of clickbait-sensational or misleading headlines designed to maximize engagement-undermines the reliability of shared information. The existing studies focus on news clickbait detection from English content using NLP techniques. To the best of our knowledge, this study is novel to focus on news clickbait detection from Urdu language content. We propose to use state of the art deep features including sentence embeddings to be applied as input to deep learning models. The dataset is prepared from authentic online source, labelled by domain experts, and pre-processed using standard steps. In contrast, traditional models, including machine learning and ensemble learning, utilize textual features and word embedding features are used as baseline models for comparing the performance of the proposed deep learning approaches. All models are evaluated using standard performance measures, including accuracy, precision, recall, F1-score, and ROC curve analysis, to determine their effectiveness in identifying clickbait in Urdu news headlines. The results show that the Bi-LSTM model with sentence embeddings achieved the highest accuracy of 88% for clickbait identification in low resource language.