Netflix Unveils Impressive AI-Driven Voice Search Feature

This is Lowpass by Janko Roettgers, a newsletter exploring the dynamic intersection of tech and entertainment, delivered weekly to subscribers.

A select group of Netflix users has a fresh way to discover movies and TV shows. The streaming giant has begun testing an AI-driven voice search feature, allowing subscribers to press the Netflix button on their remote and request viewing recommendations.

After hitting the button, users see a range of search prompts, such as “I need a good cry,” “watch in the background,” and “help me stay awake.” Each suggestion leads to tailored viewing options, and there’s also an “Ask” button featuring a waveform icon. Tapping it initiates an AI-powered voice search that responds to natural language queries with relevant recommendations.

Having tried out this voice search feature for a few days, I was impressed for two main reasons: even in its beta phase, it effectively delivers intriguing recommendations for a wide array of requests. It also bypasses the voice assistants and search functions integrated into smart TVs and streaming devices, showcasing the ongoing competition between TV operating systems and the streaming services they host.

Currently, Netflix is testing this voice search with a limited number of users in the U.S. I accessed it on a Chromecast with Google TV and a TCL Google TV, but it wasn’t available on Roku or Fire TV devices. Some users with multiple devices running the same smart TV OS reported that they could only access the feature on select devices.

During my testing, I found the feature adept at generating relevant suggestions for a variety of prompts, from the straightforward (“date night movies from the ’80s” led to The Breakfast Club) to the niche (“I like the music of Brian Eno, what should I watch” resulted in Abstract: The Art of Design).

I pushed the feature further than most would in everyday use, yet it consistently provided solid recommendations for unusual requests. For instance, I was surprised when asking for “fun kids TV shows about death” yielded A Series of Unfortunate Events and Raising Dion, both fitting choices.

When I said, “I had too much coffee today, what should I watch,” Netflix suggested a “laid-back” Sheng Wang comedy special and the Headspace Guide to Sleep. A more abstract prompt, “hurt people hurt people,” prompted the app to respond with, “Hurt people do hurt people. Here are some of the stories behind that,” followed by recommendations for 13 Reasons Why and It’s Okay to Not Be Okay.

Interestingly, the voice search doesn’t seem to tap into Netflix’s personalization engine just yet. When I asked for recommendations based on my viewing history, the app replied, “We can’t answer that one yet, but we’re working on it!” All responses are displayed as text, and the feature doesn’t utilize voice output, which I found perfectly acceptable.

Responses often come with suggestions to refine your search, like “more unhinged” or “more bittersweet.” However, this narrowing process doesn’t always yield the best results. For example, a search for Cantonese comedies produced relevant options, but when I followed the suggestion to filter by “more quirky,” it bizarrely recommended BoJack Horseman as a “Cantonese comedy with a quirky twist.”

There are some hiccups with Netflix’s voice search. When I asked for Blaxploitation movies, the system misheard me and responded to “Black exploitation movies,” ultimately providing no results, which was probably for the best. A request for “TV shows about porn” was also declined, but asking for “steamy TV shows” resulted in plenty of suggestions. This contrasts with Google’s Gemini assistant on Google TV, which declined the request and added, “I can help you find appropriate shows if you’d like.” This highlights why Netflix might prefer its own voice search rather than relying on the one provided by TV or streaming device manufacturers.

Another key reason for Netflix to develop its own voice search is to keep viewers engaged within its app, exclusively recommending Netflix content. In contrast, smart TV OS operators like Google, Roku, and Amazon prefer a universal search that includes results from their own services and those of partners they can monetize.

This ongoing power struggle between platforms and content providers has been a long-standing issue, particularly in voice search. For users, this often leads to confusion: pressing the mic button on your remote while using Hulu or Disney Plus typically accesses the universal search function, mixing results from the current app with those from other providers. However, doing the same on Netflix or YouTube restricts the search to their respective catalogs.

Thanks to their significant market presence, YouTube and Netflix have secured voice routing privileges that smaller publishers lack. Initially, they used this privilege for basic speech-to-text functionality, allowing voice queries to be processed in the same search fields accessible via remote.

With the rise of large language models, both platforms are now leveraging these privileges for more comprehensive voice assistance. YouTube recently introduced its own conversational AI tools, enabling viewers to access Gemini-powered voice features on devices from Amazon, Roku, and others.

Similar to YouTube’s approach, Netflix’s AI voice search demonstrates the benefits of apps having mic access. It shouldn’t require massive market share or a dedicated button on your remote for publishers to create such experiences.