Sunday 4 October 2026
Afternoon EditionRSS
ZOTPAPER

News without the noise

ZOTPAPER
Tech

YouTube Lets Users Build Custom AI-Powered Feeds

New feature uses Gemini to generate personalized video recommendations from user prompts

By LineZotpaper
Published
Read Time2 min
Sources2 outlets
YouTube announced Wednesday a new feature called custom feeds that lets users describe the kind of videos they want to see, using Google's Gemini AI model to generate a personalized feed pinned to the top of the home page. The move follows similar efforts by other social platforms, including Bluesky, to give users more control over their recommendation algorithms.

YouTube has introduced a new feature called custom feeds that allows users to build their own curated video recommendations by typing a description into a prompt box. The feature, announced Wednesday, leverages Google's Gemini AI model to interpret the request and generate a feed tailored to the user's needs. For example, users could request a feed of "video podcasts to watch during a 30-minute train commute," or videos for a specific mood, such as "relaxing commentary videos to unwind with."

The resulting feed appears as a tab pinned to the top of the YouTube home page. Because the creation process is AI-powered, users can provide detailed instructions about what to include, exclude, or prioritize.

YouTube's announcement follows a broader trend among social networks toward user-controlled algorithms. The decentralized social network Bluesky popularized the concept of custom feeds years ago as a way to differentiate itself from rivals like X. Earlier this year, Bluesky introduced an AI tool called Attie that further simplifies feed creation. Other services have also rolled out similar tools, allowing users to shape their content experience through natural language prompts rather than relying solely on platform-designed algorithms.

§

Analysis

Why This Matters

  • It gives YouTube users more direct control over their video recommendations, potentially reducing frustration with default algorithms.
  • The shift toward user-described feeds could change how content is discovered and consumed on the platform, affecting creators' visibility.
  • It signals that major platforms are increasingly adopting AI-powered personalization tools that let users set the terms, a trend that may spread to other services.

Background

Social media platforms have long faced criticism over opaque, one-size-fits-all recommendation algorithms that can lead to echo chambers or promote unwanted content. In response, some newer platforms like Bluesky introduced custom feed features that allowed users to choose or build their own curated streams. The idea is gaining traction as advances in AI, such as large language models, make it easier for users to describe their preferences in natural language rather than manually filtering content.

Key Perspectives

YouTube users: Gain a tool to tailor feeds to specific interests, moods, or time constraints, potentially improving the viewing experience and reducing time spent scrolling through irrelevant videos. Platform owners: Adopting user-defined feeds may increase engagement and satisfaction, but it also reduces the platform's control over what content is promoted, which could affect advertising revenue and content moderation strategies. Critics/Skeptics: Concerns include whether AI-generated feeds will accurately reflect user intent, possible biases in the Gemini model, and the potential for the feature to be gamed by bad actors to amplify certain content. There may also be privacy implications from the detailed user descriptions.

What to Watch

  • User adoption and feedback: Will custom feeds become a widely used feature or remain a niche tool?
  • How YouTube handles content moderation within AI-generated feeds — for instance, whether it filters out policy-violating videos from custom feeds.
  • Competitor responses: Whether other major platforms (e.g., Instagram, TikTok) introduce similar prompt-based feed customization.

Sources

Zotpaper

Written by software from the reporting listed above, scored by an automated standards desk, and published without a person reading it first. If something here is wrong, tell the editor and it will be put right.

How we workReport an errorSubscribe