First impressions are key, especially in the crowded world of digital content on platforms like YouTube, where thumbnails play a vital role in grabbing viewer attention. This paper investigates the impact of thumbnail presentations on user engagement and bias within YouTube’s recommendation algorithms. Our study systematically examines the attributes of thumbnails, such as colorfulness, brightness, and image quality, and their potential associations with user engagement metrics. Additionally, we delve into the analysis of thumbnail captions using advanced image captioning models, exploring how accurately these descriptions reflect the content and influence viewer behavior. From these analyses, we infer that attributes like brightness, colorfulness, or the quality of thumbnails do not significantly influence engagement metrics for these narratives. Our findings indicate that thumbnails featuring content with a universal appeal or related to universally engaging themes such as mystery, fantasy, science, and military subjects are more likely to attract higher viewership. This research not only offers a comprehensive assessment of thumbnail effectiveness on viewer engagement but also provides insights into the biases perpetuated by YouTube’s algorithms. Through our findings, we aim to enhance understanding of content strategy on digital platforms and propose design practices that can mitigate bias and improve user interaction, thereby contributing to the field of social computing and digital media studies.