YouTube Custom Feeds Offer Control With Boundaries

YouTube announced Custom Feeds on Wednesday, giving US viewers a way to build recommendation tabs by describing what they want in ordinary language.

A viewer might request DIY channels, short cooking videos or podcasts suitable for a 30-minute commute. Large language models interpret the prompt, and YouTube assembles a feed that refreshes with videos fitting the request. Multiple feeds can be saved, while the familiar Home page remains available through the “All” tab. The company plans to offer the feature on mobile, desktop and TV.

Emily Moxley, YouTube’s vice president of product management, told Wired during a demonstration that users regularly ask for more control. “With custom feeds, we're giving them a natural-language way that they can do that,” she said.

That description is accurate within a carefully drawn boundary. Custom Feeds let viewers request another set of recommendations. They do not provide controls for changing the objectives, signals or ranking rules behind YouTube’s main recommendation system.

The Algorithm Gets a Guest Room

YouTube explored a more direct approach before settling on separate tabs. The product team tested a feature that allowed viewers to steer their overall recommendation algorithm, according to Wired, then abandoned it because test users feared damaging a feed trained on years of viewing.

“We found users were nervous and afraid to do that,” Moxley said. “They were worried about messing it up.”

Custom Feeds solve that product problem by putting experimentation in a separate workspace. Someone can request videos about college football this week without deliberately rebuilding the Home feed around that passing interest. The design also reduces the perceived cost of a bad prompt: if the resulting tab is dreadful, the regular feed is still one click away. An algorithmic mulligan is easier to sell than algorithmic surgery.

The separation has a leak, however. TechRepublic reports that videos watched through a custom feed can enter a viewer’s standard watch history and may influence Home recommendations. A tab can therefore isolate the initial request while the subsequent viewing behavior travels back into YouTube’s broader system.

Taken together, those mechanics describe bounded control. The viewer chooses a temporary context and YouTube retains authority over ranking within it. Watching the results may then train the wider system through the same behavioral signals Custom Feeds were designed to supplement.

That conclusion has limits. YouTube has not disclosed in these reports how strongly custom-feed viewing will affect Home, how long any effect might last, or whether different interactions carry different weight. A single click could have little visible impact, while sustained viewing could have more. Claims that the feeds are either fully isolated or certain to contaminate Home would outrun the available information.

From Inferred Taste to Stated Intent

Custom Feeds did not appear fully formed at this week’s event. Android Police documented an experimental “Your Custom Feed” option in April 2026 that regenerated the homepage from text prompts. The announced version keeps that core function, shortens the name and moves toward a broader US release.

That history clarifies what changed. Recommendation systems have long inferred preferences from behavior such as viewing history, likes and dislikes. The April experiment added a second input: a sentence stating the viewer’s current intention. Instead of spending an evening clicking woodworking videos until the system catches the hint, a user can ask for beginner furniture projects under 15 minutes.

Stated intent can convey combinations that behavioral signals capture slowly. It can include subject, length, mood and situation in one request. It can also be misunderstood by the language model or interpreted through YouTube’s existing catalog and ranking priorities. The prompt supplies instructions, not an inventory of every eligible video or an explanation of why each result won.

The feature also sits inside a much larger AI package. At Made on YouTube, CEO Neal Mohan said the company had launched more than 3,000 product updates and improvements since the start of 2026. The Hollywood Reporter’s account lists creator tools for suggesting video structure, testing thumbnails and editing through conversational commands, alongside expanded follow-up questions in Ask YouTube.

Those additions connect production, discovery and optimization inside the same platform. Creators receive tools to make and test content; viewers receive tools to describe what they want; YouTube remains the broker deciding which videos meet the request. Niche creators could benefit when prompts surface work that a broad Home feed overlooks. No published rollout data yet show which channels gain exposure, whether established creators retain their advantage, or how diverse the generated feeds will be.

Prompting is Becoming the New Settings Menu

YouTube is joining a broader experiment in conversational personalization. Meta introduced Your Algorithm on Instagram to let people adjust the posts they see, while Spotify released Prompted Playlists for customized listening, Wired reported.

All three replace at least some menu navigation and behavioral training with a request written in natural language. Their objects differ: Instagram adjusts a stream of posts, Spotify assembles music, and YouTube creates separate video tabs while preserving Home. The comparison therefore shows a common interface trend, not proof that the underlying recommendation systems work alike. The available descriptions do not establish whether Meta and Spotify isolate prompted activity from their wider user profiles in the way YouTube’s tabs appear to do at the interface level.

The attraction for users is straightforward. Language can express “rainy-day café jazz without commentary” faster than a row of preference toggles. The attraction for platforms is also visible. Better-matched recommendations can keep people watching, listening or scrolling, although Custom Feeds has no published retention results yet.

YouTube’s adjacent Ask YouTube feature makes that commercial context clearer. The tool can assemble product-comparison tables and answer follow-up questions from a video’s watch page. A YouTube spokesperson told CNET that the company expects Ask YouTube to keep watch time steady or improve it, adding that AI-generated chapters and summaries had helped retain viewers. Those claims concern other features and come from YouTube, so they do not demonstrate that Custom Feeds will increase viewing. They do show that user convenience and platform retention can occupy the same product roadmap quite comfortably.

For viewers evaluating Custom Feeds, three questions will reveal more than the “control” label. Does the system follow compound prompts accurately? Does activity in one custom tab noticeably reshape Home? And does the feature broaden the creators a viewer encounters, or rearrange familiar recommendations beneath a fresh coat of chatbot paint?

YouTube has made the recommendation machine easier to address. Whether it has made that machine easier to direct will become visible after viewers leave a custom tab and see what follows them home.