YouTube’s experiment in handing viewers a little more control over its recommendation machinery began in May, when it introduced a feature that let people describe the subjects, interests or moods they wanted in a custom video feed.¹

On Wednesday, the company broadened that idea. Its new Custom Feeds tool lets viewers type a plain-language prompt to create a tailored stream: perhaps video podcasts for a 30-minute train journey, or “relaxing commentary videos to unwind with.” Google’s Gemini model turns those instructions into a feed, which is pinned as its own tab at the top of the YouTube home page.²

The pitch is not that YouTube’s algorithm is being replaced, but that it is becoming easier to sidestep its all-purpose default when a viewer has a more specific need. Users can spell out what to include, what to exclude and what to prioritize—an approach that echoes feed-building tools rolled out by Bluesky, X, Meta’s apps and Spotify.²

YouTube argues the feature is especially useful because of the sheer scale of the platform. “There’s over 20 billion videos in the YouTube corpus, so you’re always one search away from a treasure trove of topics,” Emily Moxley, YouTube’s VP of Product Management for Viewer AI, said ahead of the company’s Made On YouTube event.²

The next step is multiplicity. Rather than saving one AI-made channel through the catalog, viewers will be able to create and retain several feeds on their home screens. The expansion is due to reach US users on web, mobile and TVs soon; YouTube separately said multiple-feed creation would begin rolling out on web and mobile next month.¹ The main recommendations page will remain in place, leaving users to switch between YouTube’s broad algorithm and narrower, self-described alternatives.²