YouTube Custom Feeds Lets Users Build Personalized Algorithms With AI
YouTube Custom Feeds use AI to help users create personalized video recommendations based on their interests, preferences, moods, and viewing needs.
YouTube is changing how users control the videos they see on its platform. The company introduces YouTube Custom Feeds, a new AI-powered feature that lets users describe the type of content they want to watch and create a personalized feed around that request. Instead of relying only on YouTube’s existing recommendation system, users can now tell the platform what they want in their own words. YouTube uses Google’s Gemini AI model to understand the request and create a custom feed based on the user’s preferences.
Introducing AI-Powered YouTube Custom Feeds
YouTube is rolling out a new Custom Feeds feature, which provides yet another option for users to aggregate the type of content they want to view. The user enters a request into a prompt box that describes what they’re looking for. It could be a certain topic, style, vibe, or combination of them.
For instance, a user could make a request for 30-minute train ride video podcasts; another might ask for relaxing commentary videos that help wind down after work. There are additional options to specify items to the user, such as where the feed should avoid putting information.
Gemini builds the personalized feed out of those natural-language instructions. When complete, a custom feed is displayed as a tab of its own along the top of the YouTube homepage. The new feed will give people a more immediate way to share how they’d like to see YouTube results instead of relying on past watch history and actions.
How YouTube Custom Feeds Work
Creating a custom feed starts with a simple description. Users type what they want to see, and YouTube uses AI to turn that description into a personalized collection of videos. The system can understand more detailed requests than a traditional search query. For example, users can combine different topics, specify a preferred video format, or describe a particular mood.
This makes YouTube Custom Feeds different from simply searching for a keyword. A search generally returns results related to a specific query, while a custom feed is designed to continue providing recommendations based on a broader set of instructions.
Users can also create multiple feeds for different situations. One feed can focus on educational videos, while another can feature entertainment content or videos for a specific activity. Importantly, Custom Feeds do not replace YouTube’s standard homepage. Users can continue using the main recommendation feed and switch to their custom feeds when they want a different viewing experience.
AI Gives Users More Control Over Recommendations
Related walkthrough
Flint K12 Review | How AI Is Transforming Personalized Learning for Schools
YouTube already uses a recommendation system that learns from signals such as watch history, searches, likes, subscriptions, and viewing behavior. Its goal is to identify relevant content for individual users and provide personalized recommendations.
Custom Feeds add another layer to this process because users can directly describe what they want. This approach gives viewers more control without requiring them to understand how recommendation systems work. Instead of adjusting technical settings, users can simply explain their preferences using everyday language.
For example, someone interested in technology can request videos about AI research, software development, and robotics. Another user can create a feed around short cooking tutorials or travel videos for a specific destination. The feature can therefore make YouTube’s large video library easier to navigate when users have a particular viewing goal.
YouTube Does Not Replace Its Main Algorithm
The new feature does not mean YouTube completely replaces its existing recommendation algorithm with a user-controlled system. The main YouTube homepage continues to provide personalized recommendations based on the platform’s existing systems. Custom Feeds work alongside that experience and give users separate feeds for specific interests or situations.
This distinction is important because users may want different types of content at different times. A person may want general recommendations during one session and a focused feed for studying, exercising, commuting, or relaxing during another. YouTube also says the platform contains more than 20 billion videos. Custom Feeds can help users navigate this large collection by giving them a more specific way to request the content they want.
Custom Algorithms Become a Growing Trend
YouTube’s move follows a wider shift toward giving people more control over recommendation feeds. Other platforms also experiment with ways for users to shape the content they receive. Bluesky has offered custom feeds, while its Attie project uses AI to make feed creation easier. Meta’s Threads and Instagram, as well as Spotify, also introduce features that allow users to influence their recommendations.
The growing use of AI makes this process easier because people can describe their preferences naturally instead of manually selecting numerous settings. For platforms, these tools can also provide another way to help users discover content across large libraries. For viewers, they offer more control over what appears in specific viewing situations.
When Will YouTube Custom Feeds Be Available?
YouTube’s support for creating multiple custom feeds will be available starting next month on its website and mobile app. The company is unveiling the feature as part of its larger 2026 Made on YouTube updates, which include a handful of AI-powered tools both for viewers and creators. The feature is a new way of personalizing the experience.
Rather than relying on YouTube’s algorithm to make sense of all our preferences, the company now turns to us to tell it. As artificial intelligence becomes more embedded into recommendation engines, feeds based on personalization can be more flexible. Someone can define a goal, mood or interest and AI can take that input and convert it into a flow of what’s relevant. In the case of YouTube, it’s an additional form of interaction with its enormous video catalog (one that complements the current recommendation experience).
