YouTube has announced the most significant overhaul of its recommendation system in recent years: the company unveiled a custom feeds feature that will allow users to shape their own content feed using plain text prompts. The announcement was made through YouTube's official blog and picked up by Ukrainian and Russian media outlets. According to RBC-Ukraine, testing will begin in the US in the near future, with a full rollout planned for October 2026. New topic-based tabs will appear on the home screen of the mobile app, web version, and Smart TV, while the standard feed with general recommendations will keep its place — users will not be forced to "replace" the familiar interface.
Prompts Instead of Settings: How the Gemini Integration Works
The key difference between the new approach and previous manual filters is the absence of complex configuration panels. Users simply need to enter a text query describing the content they want. Thanks to the integration with the Gemini model, the query can be as detailed as possible: from "tips for morning workouts" to highly specific phrasings like "podcasts about technology for a 30-minute train ride." The prompt can specify topics, mood, preferred video formats, or, conversely, exclude unwanted content. In this way, YouTube effectively shifts part of the curation responsibility onto the viewer, turning the recommendation algorithm into a tool controlled by natural language.
Serialized Shorts, GIF Comments, and Voice on TV
Apart from the main announcement, the company presented a package of practical features for everyday use. Creators can now combine short videos into sequential episodes — so-called serialized Shorts — with cover images and sequential playback support. In the fall, the mobile app will gain a built-in GIF animation library for replying to comments. On TVs, viewers will be able to dictate comments by voice using a text-to-speech tool. The Ask YouTube feature, previously focused on search, has now learned to compare products shown in videos, provide specification tables, and answer questions about items in the frame — for example, about a blogger's outfit, even if it is not mentioned in the video.
YouTube Music and Podcasts: AI Tools for Subscribers
The updates also affected the music service. YouTube Music subscribers will get an Ask Music feature, allowing them to request additional information about artists and manage the playback queue through a dialogue with the AI. Separately, the Podcast Lineup tool is launching, which weekly compiles a short audio preview of recommended podcasts — essentially a personal "radio digest" assembled by the algorithm based on listening history.
Streaming: Coachella Through 2030 and Auto-Dubbing in 2027
In the live streaming direction, YouTube extended its exclusive rights to broadcast the Coachella music festival through 2030 and announced the React Live and Go Live Together features, enabling bloggers to host joint live streams. The most ambitious project in this area is the Live auto-dubbing pilot tool, scheduled to launch in early 2027. It will provide automatic dubbing of live streams in real time; in the first phase, support will be limited to English and Spanish.
Contradictory Data
No significant discrepancies in figures or dates were found between sources. All four verified outlets (RBC-Ukraine, 24tv, Itzine, IT-Center) reference the same primary source — the YouTube Official Blog — and convey an identical timeline: testing in the US "in the near future," full rollout in October 2026, and the auto-dubbing pilot in early 2027. The only minor difference is in the wording: some outlets state that GIF comments will appear "in the fall," whereas the original announcement refers to the mobile app without specifying a particular month. This is not a contradiction, but a difference in the level of detail.
Overall, the announcement demonstrates a strategic shift by YouTube from a "black box" of recommendations to a transparent, user-controlled interface. The integration of Gemini as a language layer and the expansion of AI tools into music, podcasts, and streaming indicate that the company intends to make artificial intelligence not just a filter, but a full-fledged "editor" of the user experience. For advertisers and creators, this means that content personalization will become even more granular, and the competition for viewer attention will shift from the "the algorithm showed it" plane to the "the viewer chose it themselves" plane.
