Netflix began rolling out an AI Discover tab on Tuesday, its most direct attempt yet to replace endless scrolling with a conversation about what someone actually wants to watch. The first feature inside the tab takes a stated mood, available time, viewing history or the current trending list, and returns a set of themed collections built from clips of shows and films.
The themes are deliberately conversational rather than genre-based: something light, scare me, family movie night, quick watches and others in the same register. Picking one narrows the recommendations, and a user can keep drilling down or hit "something different" to start over. Every title carries a "Tune Your Recommendations" button next to Play and My List, which lets viewers save a suggestion they liked or block a title from being proposed again.
Netflix describes the tab as a test kitchen where it will try out new recipes for discovery, and says feedback gathered there will decide which features graduate into the main experience. Underneath sits a new intelligence layer the company says responds to what users say in real time and keeps learning — a description that leaves open whether recommendations are generated per request by a language model or assembled from a ranking system steered by natural-language input.
The rollout is narrow. AI Discover starts in the United States for members using Netflix in English on supported televisions, where a Discover button with a red gem icon and a New label appears in the top navigation. The company says more users will get it in the coming weeks, with no date for other markets or for mobile and web.
The obvious commercial motive is retention. Netflix's catalogue has grown faster than its interface can surface, and the cost of a subscriber who opens the app, scrolls for several minutes and closes it is measured in churn. Discovery surfaces are also where advertising inventory and promotional placement live, which makes a tab that learns intent more valuable than it first appears.
The design also quietly changes how recommendation quality is measured. A conventional row of titles is judged on clicks; a mood prompt is judged on whether the session ends in a play, and the Tune button creates an explicit signal for rejection rather than mere non-selection. If Netflix publishes anything about how those signals perform, it will be a more useful data point than the feature itself.
What the launch does not include is any claim about generative output. Nothing in the tab synthesises video or writes descriptions of scenes; the clips in each theme are pulled from existing titles. That restraint is consistent with how the major streaming services have approached AI so far — recommendation and production tooling first, generated imagery on screen later, if at all.
Comments (0)
Log in to join the discussion
Log InNo comments yet