YouTube Custom Feeds: Personalize with Gemini AI

TL;DR: YouTube is launching ‘Custom Feeds’ powered by Gemini AI, allowing users to curate personalized video streams via natural language prompts. This shift toward user-controlled curation aims to enhance satisfaction in a platform with over 20 billion videos.

Key facts

• YouTube launched ‘Custom Feeds’ on September 23, 2026, during its ‘Made on YouTube’ event.
• The feature uses Google’s Gemini AI to generate personalized video recommendations from natural language prompts.
• Generated feeds appear as pinned tabs on the homepage alongside the standard algorithmic Home feed.
• Emily Moxley, VP of Product Management for Viewer AI, stated the tool lets users steer their algorithms via text.
• Custom Feeds are rolling out to all U.S. users across web, mobile, and TV platforms in fall 2026.

YouTube introduces ‘Custom Feeds’ powered by Gemini AI

YouTube announced on September 23, 2026, that it is launching a new feature called “Custom Feeds,” which allows users to generate personalized video recommendations using natural language prompts [1][2]. Unveiled during the company’s “Made on YouTube” event, this tool leverages Google’s Gemini AI model to interpret user requests and create dedicated recommendation feeds [3][6]. The initiative represents a significant shift in how viewers interact with content on a platform hosting over 20 billion videos [6].

How Custom Feeds work

The feature introduces a prompt box where users can describe the types of videos they wish to see, such as video podcasts to watch during a 30-minute train commute or relaxing commentary videos to unwind with [1][6]. The Gemini AI model processes these inputs to curate a specific set of videos that match the user’s stated preferences [2][6].

Once generated, each custom feed appears as a separate tab pinned to the top of the YouTube homepage [1][2]. These tabs sit alongside the standard Home feed, which remains untouched and continues to operate on its traditional algorithmic logic based on watch history and engagement signals [3]. This design ensures that users can switch between their curated interests and their general browsing without losing either experience [6].

Emily Moxley, YouTube’s VP of Product Management for Viewer AI, told WIRED that users want more control and more ability to steer their algorithm, and that Custom Feeds gives them a natural-language way to do that [3]. The feature moves away from reliance solely on passive signals like likes or watch history [3][6]. The goal is to help users navigate the vast volume of content more effectively by allowing them to define their own interests explicitly rather than inferring them from past behavior [1].

Industry-wide shift toward user-controlled curation

YouTube’s launch follows a broader trend among social networks and streaming platforms integrating AI-driven feed customization [1][2]. Competitors have recently introduced similar features that allow users to curate their algorithmic experiences through descriptive prompts:

• Bluesky: Launched “Attie,” a tool for building custom feeds.
• Meta: Introduced AI-powered curation in Threads and Instagram.
• X (formerly Twitter): Added options for user-defined feed filtering.
• Spotify: Enabled users to generate playlists via natural language queries [1][2]

This industry movement signals a transition from passive consumption, where algorithms decide what viewers see based on hidden metrics, to active feed construction, where users take direct control of their content streams [6]. By giving viewers the ability to steer their own algorithms, platforms aim to increase user satisfaction and engagement in an increasingly crowded digital landscape [3].

Context: YouTube’s evolving algorithm and creator tools

The announcement comes amid significant changes to YouTube’s core recommendation systems in 2026. The platform has moved toward “satisfaction-weighted discovery,” a strategy that prioritizes viewer satisfaction and retention quality over raw watch time metrics [4][7]. This shift reflects a broader industry effort to combat user burnout and improve the long-term health of viewing habits by focusing on the quality of engagement rather than just duration [5].

Additionally, YouTube continues to expand AI integration for content creators. In January 2026, CEO Neal Mohan announced that creators would soon be able to generate Shorts using their own AI likeness, a capability designed to streamline production while maintaining authenticity [8]. The introduction of Custom Feeds complements these creator-focused tools by enhancing the viewer experience, creating a more cohesive ecosystem where both sides benefit from advanced AI capabilities [6].

Rollout plans

nYouTube plans to roll out Custom Feeds to all U.S. users across web, mobile, and TV platforms in fall 2026 [3][6]. The company has not yet specified an international release date or details on potential monetization models for custom feeds [1]. As the feature enters production, it will be closely watched as a test case for how well natural language interfaces can reshape content discovery at scale [2].

Sources

1. YouTube will let you build your own algorithm with AI | TechCrunch (techcrunch.com) — 2026-09-23
2. YouTube will let you build your own algorithm with AI | TechCrunch (techcrunch.com) — 2026-09-23
3. YouTube’s Custom Feeds Give You More Control Over the Algorithm (www.wired.com) — 2026-09-23
4. YouTube’s Custom Feeds Let You Tell Gemini Exactly What You Want to Watch (www.gadgetreview.com) — 2026-09-23
5. 2026 YouTube Algorithm Updates: Month-by-Month Guide — AIR Media-Tech (air.io) — 2026-09-08
6. 2026 YouTube Algorithm Updates: Month-by-Month Guide — MilX (milx.app) — 2026-07-29
7. What’s Actually Working on YouTube in 2026 (7 Algorithm Shifts You Need to Know) — Gillian Perkins (www.gillianperkins.com) — 2026-06-15
8. YouTube will soon let creators make Shorts with their own AI likeness | TechCrunch (techcrunch.com) — 2026-01-21