Unscrupulous content creators on YouTube employ deceptive techniques such as spam and clickbait to reach a broad audience and trick users into clicking on their videos to increase their advertisement revenue. Clickbait detection on YouTube requires an in depth examination and analysis of the intricate relationship between the video content and video descriptors title and thumbnail. However, the current solutions are mostly centred around the study of video descriptors and other metadata such as likes, tags, comments, etc and fail to utilize the video content, both video and audio. Therefore, we introduce a novel model to detect clickbaits on YouTube that consider the relationship between video content and title or thumbnail. The proposed model consists of a stacking classifier framework composed of six base models (K Nearest Neighbours, Support Vector Machine, XGBoost, Naive Bayes, Logistic Regression, and Multilayer Perceptron) and a meta classifier. The developed clickbait detection model achieved a high accuracy of 92.89% for the novel BollyBAIT dataset and 95.38% for Misleading Video Dataset. Additionally, the stated classifier does not use meta features or other statistics dependent on user interaction with the video (the number of likes, followers, or comments) for classification, and thus, can be used to detect potential clickbait videos before they are uploaded, thereby preventing the nuisance of clickbaits altogether and improving the users streaming experience.