YouTube will let you ask AI to make a custom video feed

YouTube has announced a new AI feature whose core concept is to let users rely not only on the existing recommendation algorithm, but to describe in text the type of videos they want to watch and ask AI to generate a set of customized video feeds. According to the announcement cited by The Verge, these custom content feeds can be built according to a user's specific interests, current mood, or favorite topics. For example, a user might not just search for a single keyword, but express the direction of content they want to explore at the moment in a more natural way that is closer to a description of their needs. YouTube then builds a corresponding video feed based on that description. Another key point of this feature is that, once generated, the custom feed can be pinned to the top of the YouTube homepage, letting users turn certain long-term interests or short-term situational needs into a fixed entry point. For ordinary viewers, this may reduce the cost of switching back and forth among homepage recommendations, search, and subscribed content; for creators and marketers, it also means that YouTube's content-discovery mechanism may place more weight on interest sets that users express in natural language, rather than only on a single video's title, keywords, or viewing history. The original piece does not currently explain the name of the model behind it, the full rollout regions, supported languages, whether it is open to all accounts, or how custom feeds will affect existing recommendation ranking, so it is unwise to extrapolate technical details or business impact. In terms of product direction, however, this is yet another step by YouTube in putting generative AI into the content-discovery interface, with the goal of partly shifting homepage recommendations from passive reception toward letting users actively define their viewing context.