This paper presents results from an inquiry on how different digital platforms are used to communicate about and engage with climate politics, at the occasion of the Paris climate change conference, COP2. Our focus will be on Youtube, and to a lesser extent Twitter. Among the different platforms, Twitter stands out as the one with the most activity. Interestingly, more than half of the publications about COP21 contained a link. 3% of these link to YouTube videos. Yet the collection of COP21 YouTube videos shared by users on Twitter, and the results from queries on COP21 curated by the Youtube algorithms barely overlap. So, how do algorithms organize issue exploration in Youtube? Looking at visualizations of “up next” video connections among query results, something particularly interesting is made visible. By far the most watched COP21 video on YouTube is a short film from a French NGO, Noé that shows Koko the gorilla signing out a message to international negotiators about the need to include biodiversity in the Paris climate agreement. Yet despite amassing over 1 million views, the Koko video is largely hidden from YouTube users who searched the website for ‘COP21’ by relevance. This presents a puzzle: how can a YouTube video be both popular and invisible? Employing YouTube Data Analysis Tools, we attempt to solve this puzzle by illuminating the role of YouTube’s ‘relevance’ and ‘related to’ algorithms in issue exploration. The invisibility of Koko to users exploring COP21 provides insights into the logics at work within YouTube; in particular, how they attempt to maximise users’ viewing time. Completing conceptualizations in terms of personalized filter bubbles, the proposed paper introduces the concept of issue drifts to show how YouTube’s ‘related videos’ feature opens up some exploration paths while closing down others on other criteria than personalization. We conclude by opening a discussion on multiple platform effects on issue exploration.