Executive Overview
YouTube has officially announced a feature called custom feeds, giving users the ability to construct their own tailored recommendation algorithms using generative artificial intelligence. Unveiled during the company's Made On YouTube event, the feature allows viewers to type plain-language descriptions into a prompt box, explicitly guiding what content should appear in specialized tabs pinned directly to their home page.
Driven under the hood by Google's flagship Gemini AI model, this addition aims to solve discovery challenges across YouTube's massive library of over 20 billion videos. Rather than replacing the primary recommendation engine, these prompt-based feeds exist alongside the main feed, granting users direct control over content curation across web and mobile platforms.
What exactly changed with YouTube's recommendation system?
Traditionally, YouTube relies on automated recommendation systems that analyze user viewing history, engagement metrics, and behavioral patterns to determine what videos to present on the main homepage. With the introduction of custom feeds, YouTube is offering an alternative layer of user-directed curation powered by prompt engineering.
Viewers can enter explicit, detailed text prompts describing the precise category, format, duration, or tone of content they want to see. Once generated, these specialized custom feeds are pinned to the top of the YouTube home page in dedicated tabs, enabling viewers to switch seamlessly between their personalized custom streams and the standard global algorithm.
Key Capabilities of Custom Feeds
- Prompt-Based Curation: Describe specific content desires in plain text via a dedicated prompt box.
- Gemini AI Integration: Leverages Google's Gemini model to interpret complex user instructions.
- Pinned Navigation: Custom feeds live as separate tabs at the top of the home page.
- Preserved Main Feed: The primary algorithm remains intact and active alongside custom tabs.
How does natural language prompting control the algorithm?
Because the underlying system is powered by Google's Gemini AI, users are not restricted to basic keyword tags or simple category dropdowns. Instead, the feature supports lengthy, highly specific text inputs where viewers can define multi-layered instructions regarding what to feature, prioritize, or exclude.
For example, YouTube suggested users could request a feed specifically for video podcasts tailored to a 30-minute train commute, or ask for mood-based recommendations such as relaxing commentary videos to unwind with. The Gemini model parses these nuanced parameters to evaluate candidate videos across YouTube's massive catalog.
— Emily Moxley, VP of Product Management for Viewer AI at YouTube
How does this fit into the broader tech industry trend?
YouTube's release follows similar moves across the broader social media and entertainment landscape. The concept of empowering users to construct custom algorithmic feeds was initially popularized by decentralized network Bluesky as a primary point of differentiation from legacy platforms. Bluesky further accelerated this model earlier this year with the introduction of its AI feed-building tool, Attie.
Since then, several major tech platforms have launched custom feed features—many utilizing AI—including Meta's Threads, Instagram, X, and Spotify. By giving users explicit controls to define what appears in their timelines, platforms are responding to user demand for customizable discovery mechanisms that operate beyond standard automated algorithms.
Quick FAQ Masterclass
Will custom feeds replace my existing YouTube home feed?
No. Custom feeds will not replace your main YouTube feed. Instead, they appear in separate tabs pinned to the top of the home page, leaving your standard primary recommendations fully operational.
What artificial intelligence technology powers this feature?
The feature leverages Google's Gemini AI model to interpret complex, detailed user descriptions and curate corresponding videos from the library.
Can I build multiple custom feeds?
Yes. YouTube announced support for creating multiple custom feeds, allowing users to build separate feeds for different moods, interests, or routines.
When will custom feeds become available to users?
YouTube stated that support for creating multiple custom feeds will begin rolling out across web and mobile platforms starting next month.
Next Steps & Future Outlook
The global rollout for custom feed creation begins next month on both web and mobile client applications. As users gain access to Gemini-powered prompt tools, YouTube aims to facilitate deeper exploration across its catalog of over 20 billion videos. By bridging direct natural language input with underlying recommendation systems, YouTube is taking a significant step toward giving viewers active agency over their digital content experience.