YouTube Lets You Build Your Own Algorithm With AI
YouTube is letting users construct custom feeds from a text prompt, powered by Gemini — the company's latest concession to the idea that one giant recommendation algorithm can't serve everyone.
Published: 2026-09-24 Category: Quick Take Sources: TechCrunch — YouTube will let you build your own algorithm with AI
The Update
YouTube says it will let you build your own algorithm — well, kind of. The company announced a feature that lets users type a description of the kind of feed they want into a prompt box on YouTube, and the AI constructs a dedicated feed around that request. You might ask for a feed of video podcasts to watch during a 30-minute train commute, or request videos for a particular mood, like "relaxing commentary videos to unwind with."
The prompt-based feature relies on Google's Gemini AI model to build the feed, which then gets pinned to the top of the YouTube home page in its own tab. Because it's AI-powered, the request can be richly detailed — users can describe what the feed should feature, exclude, and prioritize at length.
The addition follows a wave of similar moves across social platforms. Decentralized network Bluesky helped popularize the user-built-feed idea as a differentiator against X, and this year rolled out its own AI tool, Attie, to make feed-building easier. Meta's Threads, Instagram, X, and most recently Spotify have all launched feed-building features, many of them AI-powered.
Why This Matters
This is a meaningful philosophical shift for YouTube, which built its dominance on a single, opaque, hyper-tuned recommendation algorithm that has been the subject of endless criticism around filter bubbles and radicalization. Giving users the tools to override that algorithm — even partially — acknowledges that personalized-to-a-fault isn't the only valid experience. It hands a sliver of control back to the viewer, a response to growing demand for agency over feeds.
For Google, the move is also a Gemini showcase. Prompting a feed into existence is a highly visible, consumer-facing demonstration of what a large language model can do beyond chat, directly inside the most-watched platform on the internet. It positions Gemini as the connective tissue between intent and content.
The Caveats
The limits are worth naming. This is not full algorithmic transparency — YouTube's core recommendation system still runs underneath, and the AI-built feed is a curated layer on top, shaped by whatever Gemini happens to surface. The feature also presumes users know what they want well enough to prompt for it, which inverts the traditional value of discovery: the joy of YouTube was often finding what you didn't know you wanted. If custom feeds succeed, they may fracture the shared-video experience further into silos — the very fragmentation critics say social platforms are already suffering from. And as with all AI-generated feeds, there's a real question of whether users will trust a prompt-built feed more than the algorithm they already distrusted.
Sources: TechCrunch — YouTube will let you build your own algorithm with AI (Sep 23, 2026).