How Netflix AI Discover Works and What Retailers Can Learn

Netflix has started testing AI Discover, a new tab in its TV app that suggests titles based on themes such as mood, who is watching and how much time is available. The test began on November 3, 2026, with a small group of US members, but the approach matters to anyone who runs product recommendations.

Netflix announced the feature on its official newsroom, and several technology outlets covered it. The company describes the tab as a place to try new ways of helping members find something to watch. Member feedback will decide which ideas are kept, improved or moved into the main Netflix experience.

How the AI Discover tab works

Members in the test see a “Discover” button with a red gem icon and a “New” label in the top navigation of the TV app. The tab opens on a row of themes, each introduced by a short clip from a film or series, so viewers get a feel for the theme before choosing it.

Netflix’s examples include “Something Light,” “Scare Me,” “Family Movie Night” and “Quick Watches.” According to TV Tech, other themes are built from titles that are new or popular, or from what the member has watched recently.

Picking a theme brings up a list of matching titles. More specific themes then appear to narrow the list, and a “Something Different” option serves a fresh set of suggestions when nothing fits.

Every title also carries a “Tune Your Recommendations” button next to “Play” and “My List.” Members can use it to keep a title they like or ask Netflix not to suggest it again. Netflix says AI Discover will remember that choice, so rejected titles stop coming back.

What the AI does

Netflix says AI helps create the themes and explains why a title was recommended. The company also claims AI Discover runs on a new intelligence layer that responds to member feedback in real time and keeps learning. It has not said which AI models it uses or how that layer works.

Netflix frames the project as an experiment. “Think of AI Discover as our test kitchen,” the company wrote in its announcement, as quoted by TheWrap. Features inside the tab are trials and may never become a permanent part of the app.

From viewing history to viewing situation

Netflix’s recommendations have long relied on member signals such as Thumbs Up ratings and titles added to “My List.” According to TechCrunch, Netflix presents AI Discover as an additional way to explore that sits alongside those signals.

The difference lies in the starting point. Standard recommendations infer taste from past behavior, while AI Discover asks members what they need right now. The same person may want something different on a quiet night in than on a family movie night, and Netflix uses that very example to explain why it built the feature.

Netflix has tried to solve the “what should I watch” problem before. ThePrint notes that Netflix launched Play Something in 2021, a button that picked a title based on a member’s preferences, and retired it in 2023. In 2025, Netflix tested an AI-powered search on iOS that let users describe what they wanted in everyday language.

ThePrint also points to two pieces of context. A 2017 Wired report said more than 80% of what members watched on Netflix came from its recommendations. A Bloomberg report in October 2026 said subscribers were opening the app less often and watching for shorter periods than a few years earlier. Both figures come from media reports, and Netflix did not cite them when announcing AI Discover.

A limited test with open questions

For now, AI Discover is available only to a small group of US members who use Netflix in English on supported TVs. Netflix says access will widen gradually over the coming weeks.

Several details have not been announced. Netflix has not disclosed the size of the test group, the list of supported TVs, the AI models involved or a timeline for other countries. There is no announced plan for Indonesia, so viewers there cannot try the feature yet.

The claim that AI Discover makes choosing easier also remains Netflix’s own. No independent measurement of its recommendation quality has been published, and Netflix itself treats the tab as an experiment to be judged by usage and member feedback.

What this means for online retailers

Netflix’s problem is familiar to online stores and marketplaces. Large catalogs tire shoppers out, and many leave without buying. AI Discover offers several lessons that apply to retail management solutions, from small web shops to modern store chains.

First, recommendations can start from the shopper’s situation. Alongside “similar items” or “frequently bought together,” a store can offer needs-based entry points such as “gifts for mom,” “monthly essentials on a budget” or “back-to-school supplies.” Themes like these also help first-time visitors who have no purchase history yet.

Second, recommendations should be easy to correct. The “Tune Your Recommendations” button gives members control over what appears. In ecommerce, the equivalent is a “not interested” or “hide this item” option. Direct feedback of this kind has a clear meaning, which makes it more useful for improving a model than guesses based on clicks alone.

Third, explain why an item appears. Netflix says AI Discover will explain its picks. A short line such as “Shown because you viewed running shoes this week” helps shoppers understand the suggestion and helps the store’s team check that the system behaves sensibly.

Fourth, roll out in stages. Netflix opened AI Discover to a small group and called it a test kitchen. A store can do the same by showing a new recommendation model to a share of visitors, then comparing click-through rates, conversion and how often shoppers hide items. Building and testing such models is the kind of work covered by machine learning services.

Fifth, treat the data with care. Preferences, purchase history and shopper feedback are personal data. Stores should explain how they use them in a privacy policy, ask for consent where required and give shoppers a way to clear their preferences. A related lesson on product data readiness came out of Visa’s benchmark of AI shopping agents.

Frequently asked questions

Is Netflix AI Discover available in Indonesia?

No. The first test covers only a small group of US members who use Netflix in English on supported TVs. Netflix has not announced timing for other countries, including Indonesia.

Is AI Discover a chatbot?

Netflix’s announcement describes interaction through choosing themes, narrowing them and using feedback buttons, and it does not mention a chat feature. AI is used to create themes, explain recommendations and adjust suggestions based on member feedback.

How is AI Discover different from regular Netflix recommendations?

Regular recommendations infer taste from viewing history and signals such as Thumbs Up and My List. AI Discover asks members to pick a theme that fits the moment, then adjusts its suggestions from direct feedback. Both run side by side during the test.

Closing thoughts

AI Discover is still a limited experiment, but its direction is clear: recommendations that start from the current need, can be corrected and come with a reason. Online stores can adopt all three step by step without Netflix-scale technology. If you are planning product recommendations or want to test a new approach safely, the Merakit team is ready to help you design it.

Sources

• Netflix: Introducing AI Discover (November 3, 2026)
• TechCrunch: Netflix’s new AI Discover tab helps you find something to watch based on your mood (November 3, 2026)
• TheWrap: Netflix Adds AI Discover Option to TV App (November 3, 2026)
• TV Tech: Netflix Deploys Beta ‘AI Discover’ Feature (November 4, 2026)
• ThePrint: Can’t decide what to watch on Netflix? AI Discover wants to fix that (November 4, 2026)