YouTube lets you build your own algorithm with AI
YouTube’s new custom feeds let users describe the videos they want in their own words, then use Gemini to create a personalized feed around that request.
YouTube has announced a new feature called custom feeds, which enables users to shape their video recommendations by typing a simple text description of what they want to watch.
When a user submits a short description of the desired content, YouTube’s Gemini AI model analyses the request, selects matching videos, and assembles a tailor‑made feed.
This approach differs from the platform’s existing recommendation algorithms, which rely heavily on watch history and implicit signals; here the user’s intent is expressed explicitly in natural language.
For creators, the feature offers a way to reach niche audiences that are actively searching for specific topics, while advertisers can more quickly place their messages in front of viewers who have already indicated interest in related products.
Currently rolled out as a limited beta in the United States and select European countries, YouTube plans to expand the feature’s availability and further enhance Gemini’s capabilities in upcoming updates.
