In the digital age, the integrity of online media has been significantly compromised due to the prevalence of clickbait headlines, which mislead users by offering sensationalized content. The proliferation of misinformation through exaggerated or misleading headlines has eroded trust in digital platforms, thereby emphasizing the need for effective methods to detect clickbait. This study proposes machine learning (ML)-based models for detecting clickbait in news headlines, leveraging a balanced dataset of 32,000 instances of both clickbait and non-clickbait headlines. The term frequency-inverse document frequency (TF-IDF) technique is utilized for feature extraction, while the performance of advanced gradient boosting algorithms, namely light gradient boosting machine (LightGBM) and categorical boosting (CatBoost), is compared for clickbait detection and classification. The results show that the CatBoost model outperforms LightGBM with an impressive 98.61% accuracy, 98.33% precision, 98.99% recall, and an F1 score of 98.60%. In comparison, LightGBM achieved 96.20% accuracy, 97.40% precision, 96.89% recall, and an F1 score of 97.14%. This study demonstrates the superior effectiveness of gradient boosting models for detecting clickbait in news headlines, contributing to more reliable real-time detection of misinformation and ensuring the authenticity of content across digital platforms. The findings underscore the potential of these models to enhance the quality of online media and restore trust in digital journalism.