YouTube Lets You Build Custom Feeds With a Prompt
YouTube announced custom feeds on September 23, 2026, a feature that lets users describe in their own words the videos they want recommended, at its Made on YouTube event. TechCrunch
You create a feed by typing a description into a prompt box. Google's Gemini AI model then builds the feed around that request. Each feed is pinned to the top of the home page in its own tab.
The custom feeds will not replace YouTube's main recommendations feed. The main feed stays. Support for creating multiple custom feeds will roll out on web and mobile starting next month.
As examples, YouTube pointed to video podcasts for a 30-minute train commute or relaxing commentary videos to unwind with. Emily Moxley, YouTube VP of Product Management for Viewer AI, said the company holds over 20 billion videos.
The news came at YouTube's annual Made on YouTube event in New York City on September 23, 2026. The Alphabet-owned company also unveiled AI creator tools and shopping updates at the event. Reuters
The broader context here is a shift from inferred intent to declared intent. Most recommenders learn from implicit signals such as watch history, session behavior and patterns across similar viewers. A prompt-defined feed inverts that flow. You state the need, and the system must find candidates, rank them and refresh the list. The tensions are familiar: precision against recall, freshness against relevance, and sticking to the request against adding serendipity.
In my view, the lasting change is inspectability. A feed defined by a sentence you wrote is easier to reason about and revise than one shaped by months of clicks. Plain language is still brittle as a control. Narrow prompts can starve the candidate set, while broad prompts can collapse toward the default ranking. Multiple pinned tabs point to a practical fix, with users keeping several narrow feeds rather than tuning one master feed.
Looking at what this means for builders and creators, explicit feeds tie distribution to stated need rather than inferred taste. That rewards catalog depth, consistent metadata and clear formatting, because language must match inventory before ranking. Evaluation shifts too. For a declared task like filling a commute, click-through and raw watch time weaken as proxies. Task completion, repeat visits to a tab, and prompt edits matter more. If viewers adopt it, recommender work moves toward constrained retrieval, instruction following and clear controls. That would give viewers a steering wheel they have not had before, without removing the main feed for those who prefer to scroll.
