Q&A: Google's Shalini Govil-Pai Talks the Future of Android TV
Like a lot of streaming consumers, I typically go to specific streaming apps for content. Meanwhile, many TV providers have revolutionized the way consumers can find things.
In advance of my interview with Shalini Govil-Pai, Google’s VP and general manager of TV platforms, I asked Google for a demo device so I could reset my expectations for how a hardware-based streaming service should work. I was really impressed with how easy it was to find things I knew about and things I wanted to try, while offering an interesting combination of premium and social content.
In our interview, which follows—edited for space and clarity—Govil-Pai and I discussed ongoing developments at Google TV, the state and future of streaming, and the growing impact of AI on streaming recommendations, user experiences, and monetization.
Krefetz: What is your role at Google?
Govil-Pai: I manage the Android TV platform, which is our operating system used by operators, TV retailers, and OTT devices. On top of that, we have developed an interface which we call Google TV, but it has an experience where we combine signals from user interests to predict what they might want to watch. That’s been a real differentiating feature for us over the years.
Content Discovery and the User Experience
Krefetz: Recommendations have been in the industry for a long time. How does Google’s approach differ?
Govil-Pai: Recommendations are changing so much. We have a system that is based a lot on collaborative filtering and other mechanisms that we use at Google. We’re looking a lot at what YouTube and TikTok are doing. I think they’re using phenomenal models, including using LLMs that are showcasing very powerful results. Our goal is, we want you to spend your time watching, not searching. We don’t want it to become spammy. We don’t want it to become addictive either.
Krefetz: Is there anything you can share about either how the recommendations work or how the consumers respond to them?
Govil-Pai: In terms of how users respond, I would say, everyone used to go to apps directly, and then they would dig into the app to figure out what they should watch. If they didn’t like something in that app, they would pop back to the home screen, and then they would go to another app, which is clearly not a great experience.
Over the last few years, we’ve found that people love to go directly to the content from their home screen, and we have a number of different rows via which we actually recommend content. The first signal is not even recommendations; it’s knowing that you are watching an episode of a season and so, probably, you want to get back to where you were.
Also, we have what we call “Top picks for you,” and that’s based on all of the signals you’ve given us. For example, you’re a sports fan, and so sports—especially live sports—is number one on your recommendation list. Then, below the fold, we have many different roles; for International Women’s Day, for example, we’re doing a collection on powerful-women content. We did the same for Black History Month. Then we have dramas and comedies that we’ll recommend against.
Google’s Top picks for you
The recommendations go across the apps. We have deals with all of our content partners where we get the metadata of their content, and then we’ll use it as a basis for recommendations across everything that you’ve subscribed to.
Krefetz: It looks like you have solved the search problem where viewers couldn’t find where they wanted to watch something if they already owned it or already had an account. Can you talk about how you did this?
Govil-Pai: We allow users to personalize their services at device setup, but, increasingly, we are becoming smarter about which services users engage with and derive their entitlements based on usage over time.
Krefetz: Are you using the supplied metadata or an LLM to provide recommendations?
Govil-Pai: We use metadata as well as LLMs to provide recommendations. We did a recent blog post on how we’re using the Gemini model. It says, “We’re bringing enhanced overviews of top movies and shows using Gemini technology. These overviews include full summaries, audience reviews and season-by-season breakdowns so you can make the perfect choice about what to watch.”
Enhanced overviews on Google TV
Krefetz: Is the metadata that’s being supplied by streaming services for universal search using the Nielsen/Gracenote taxonomy, or do you have your own taxonomy?
Govil-Pai: We use the standard metadata ingestion infrastructure that Google search provides and apply models, privacy, and security policies as applicable for the TV surface.
Krefetz: We are a technical publication, and I’m sure people would be interested in the most common metadata fields. Can you give me some examples?
Govil-Pai: Any attributes related to a show or a movie (e.g., title, cast and crew, ratings) are considered metadata in our infrastructure. There are hundreds of other attributes we incorporate to provide the best, most personalized recommendations for each user.
Krefetz: How has the development of AI changed how people view content?
Govil-Pai: There’s been an explosion of content. It used to be that the EPG [electronic program guide] was enough. Now, with the influx of content across Netflix, HBO, and Disney, but also through YouTube and TikTok, the content amount has grown so substantially that the old ways of viewing are no longer possible across it. That’s why over the last few years, you’ve seen everyone focus so much on recommendations that are AI/ machine learning-based. Because if you don’t have that, it’s almost impossible to find things that you really like or discover new things. You can go into a silo and just be on one particular channel, but you’ll never know what else is out there. These recommendations become a really important ingredient for everyone—both users and platforms—to be able to showcase all of the different types of content that we think you’ll like.
Over the next few years, the amount of content is going to explode even more. We’re going to see a lot more, as generative AI helps you create content even faster. I think there’s going to be even more content that’s going to be available on a global level. I think that these recommendation systems are going to have to get even smarter and better. We’re always adapting to the new paradigms.
Ad-Based Monetization Strategies
Krefetz: Can you give me some background on monetization strategies related to ad-based services?
Govil-Pai: A lot of the monetization happens in these services. A platform is not such a big monetization umbrella, although we have some levers. Mostly, the monetization comes from the services. We currently have two models for the platform itself. With one, as soon as you log on, you see a big display screen where we show ads. Typically, we try to show ads that help you understand what’s upcoming. A lot of our content partners want to showcase something that’s premium for them. They want people to sign up, or they have goals around user engagement, so they’ll advertise there. That is a platform-based advertising system. Then we also have our FAST channels. We monetize those channels through in-stream advertising.
Krefetz: Is the advertising monetization system similar to the way the traditional cable industry used to work, where they got a certain amount of minutes per hour?
Govil-Pai: It depends on the deals that we structured. With most of them, we let them advertise, and then they just give us a share of it back. On our FAST channels, we sell some of it, and our partners sell some of it.
Krefetz: Are you supporting an ecommerce system for people to buy things online?
Govil-Pai: We haven’t got anything specific on TV just as yet. What we find is when people are watching, they’re not really in the mood for performing ecommerce. On YouTube, we’ve been experimenting with QR codes, so if you watch something and you want to buy it, you can just quickly scan it and put it on through your phone.
Amazon has done something very interesting where you can just press a button, you can just press your remote, and it puts it in your shopping cart. I think those are very interesting models. How do you not take the user away from what they’re watching and still build up an inventory that we can do later? Those are things that we’re looking at, and there’s enormous demand for it.
Merging Broadcast and Digital
Krefetz: Do you have comments on the convergence of broadcast TV and digital?
Govil-Pai: I started on Google TV 7 years ago, and when I joined, I always thought there was going to be a merge of linear content where you want some via broadcast and some through broadband. But technology has a way of getting out there and growing faster than you think. Today, even a lot of sports events are being live streamed through YouTube, Amazon, Apple, and other platforms.
From a consumer angle, you may not even know or care how an event is being broadcast, but the mechanism is becoming more and more via broadband as opposed to via broadcast.
Chatbots and Android TV
Krefetz: How does Google’s OS differ from competitors’ systems?
Govil-Pai: The OS itself is called Android TV. From an app developer perspective, you can develop for the phone and then with very minor changes, you can move it over to the TV. It’s also on autos, and we just recently announced XR—our virtual reality devices, which are also based on Android.
Google’s Android XR
From an experience perspective, it’s not a group of apps that you’re going to when you start it off; it’s really a content-forward experience that’s recommendation-based.
We have also launched a UI that’s for people who want more on-demand, YouTube-style viewing. You can get the EPG directly on the platform. We were the first people to launch this along with YouTube TV. We’ve also launched FAST channels within this interface. We have more than 100 free live TV channels.
We announced at CES this year that we’re bringing Gemini to TVs. With the new hardware and far-field microphones on your TV, you can say, “Hey Gemini, show me what the solar system is about” and engage with that content with your third grader if you happen to have one. So that’s kind of the vision for us as we bring Gemini to TVs. That’s the next frontier.
Krefetz: Can you explain more about how this works?
Govil-Pai: Our plan is to bring new AI capabilities into Google TV that use our Gemini models to make interacting with your TV more intuitive and helpful. This will make searching through your media easier, and you will be able to ask questions about travel, health, space, history, and more, and the results will surface videos for added context. The Gemini model will also create customized artwork on the Google TV screen and let you control your smart home devices while the TV is in ambient mode and get an overview of the day’s news. These features will begin rolling out later this year on select Google TV devices.
Krefetz: I saw a prototype last year where you could talk to a chatbot to help find content to view. Is that in production?
Govil-Pai: No, we’re going to be launching that later this year.
Krefetz: I wrote about voice command systems eight years ago, and at that point, I interviewed somebody who said people have a hard time talking to their TV set. That seems to have changed. What changed it?
Govil-Pai: I think it’s precision. Right now, we still don’t see people talking to the TV generally. They’re very focused: “I want to watch this show. Can you help me open it and let me watch it?” Also, they’re doing it because it actually works. Ten years ago—even as recently as 5 years ago—it was still very hodgepodge. In the last year, we’ve seen a really big quality jump and, therefore, a user jump in engagement on voice.
Krefetz: What are consumers’ current favorite features of Android TV?
Govil-Pai: It’s the recommendations and the way that it’s very quick for you to get to the content right away. “Continue Watching” is a huge driver. In fact, all of my executives of Google always have feedback for me on that feature. For people who like the EPG, we have that.
Another one, interestingly, is Google Assistant. We’re working with Gemini to bring in Gemini features. It’s powerful to not have to sit and search or scroll, but just to be able to say, “Hey Google, show me Oscars that are happening right now,” and it pulls it up for you right away, opens the right app, and gets you to the content.
State of the Google TV Ecosystem
Krefetz: What does a typical TV look like today?
Govil-Pai: Prices on TVs are coming down so much that 4Ks are not that expensive anymore. All of our OEMs that we work with worldwide still require 2K support. In the U.S., it’s mostly migrated to 4K. Every TV is smart now; the chipsets are so cheap that every TV is just plugged in and it’s a streaming TV, and most people have good bandwidth in their homes to support that.
Krefetz: Are you managing sound levels so that ad-supported apps are providing a suitable level of sound? I have heard people complain about how the audio levels in TV and streaming advertising are very inconsistent. Are you managing this so that there are no blaring, loud ads?
Govil-Pai: Sound normalization usually happens within each app that is streaming content and ads, rather than centrally through the operating system. We are committed to delivering a high-quality experience in collaboration with all of our partners, and we refer these issues to the relevant partners as they are reported.
Krefetz: What are you doing to make it easier for streaming services to onboard their apps to Google TV?
Govil-Pai: We have provided easier tools based on the Android Jetpack Compose API, which are now used by many app developers, with positive feedback. We continue to provide detailed developer support documentation and partner engineering support to help our app partners.
Krefetz: What kind of viewing stats do you track on your platform?
Govil-Pai: We refer to monthly active devices that run on our platform.
Krefetz: What kind, if any, data do you share with streaming app providers on viewing patterns?
Govil-Pai: We regularly meet with app providers and share bidirectionally to ensure that Google TV is helping them gain and retain viewers, while we gather valuable feedback to make the platform better.
Krefetz: Will consumers have the chance to build their own personalized viewing interface? How would they do this?
Govil-Pai: Personalization is a key user experience of Google TV, and we are investing a lot in this space. Google TV is designed so that the more you watch and interact with the user interface, the more personalized your home screen and viewing experience will be. In 2020, we did a blog post on how viewers can make Google TV more tailored to them.
User profiles allow viewers to enjoy their own personalized space with their Google Account. They can get personalized TV show and movie recommendations, easy access to their personal watchlist, more personalized help from Google Assistant, and more. We also offer kid profiles to make Google TV a more family-friendly experience.
Additionally, Ambient Mode helps viewers stay up-to-date even when their TV is idle, providing personalized information (such as news briefs) and recommendations at a glance.
Krefetz: Will consumers be able to control agentic workflows to better configure their preferred streaming services?
Govil-Pai: Google TV has always been a content-forward experience personalized on services that the user is entitled to; we work hard to avoid confusing upsells. Google TV allows users to select their services at device setup, but, increasingly, we are using usage patterns and content affinity to help our recommendation engine determine the actual services a user may be using versus just those selected during setup. While not quite fully agentic yet, the goal for us is to always look at technologies and determine how they can be helpful for our users.
Krefetz: Google TV shows a “hero” screen, or what some companies call a home screen ad. Mine is for basketball, but I hate watching sports. Why am I seeing this ad that is taking over two-thirds of my home screen real estate?
Govil-Pai: Our home screen banner is built on top of the same robust Google and YouTube ads systems that power the hero banner ads on billions of devices. Advertisers get all of the targeting options out of the box; however, it’s up to them how narrow or wide they want to target. We’re continuing to develop and improve our advertising experience on Google TV. This includes empowering users to easily dismiss irrelevant ads (you can long-press on an ad and click “hide this ad” or see “about this ad”) and exploring more ways to better tailor the ad experience to their interests.
Full-screen hero banner ads on Google TV
Krefetz: I am curious about the content-forward experience you mentioned. It sounds like a benefit and also a challenge for people who are concerned about privacy. How does a consumer easily manage their privacy settings on Google TV?
Govil-Pai: Google TV personalizes your experience based on your viewing and entertainment-related searches when you’re signed in to your Google Account, and you have full control over this data. You can manage your privacy settings through My Google Activity Controls and find detailed information on how your data is used in the Google TV Help Center. Google maintains strict privacy standards and does not sell your personal data or use content from apps like Drive, Photos, or Gmail for ads. You are in control of your ad experience, including disabling personalization, through My Ad Center.
Krefetz: Could you provide a link to where people can understand this further and manage it?
Govil-Pai: Google’s posted privacy policy details the types of activity information that are collected and shared across our services.
Related Articles
How are the OEM and OS battles changing the way TV is marketed and distributed, from live linear to FAST, AVOD, and all other forms of free streaming? Are we heading toward a closed society in terms of CTV, as the worldwide competition for control of TV interfaces continues and dominant OEMs consolidate their power as content gatekeepers? In this article, several leading industry figures weigh in on these questions, offer insight about what is at stake for providers and users, and discuss the current CTV landscape.
29 May 2025
The case for AI-based content localization is typically made in terms of how it streamlines localization initiatives and dramatically reduces costs, particularly for smaller publishers that might not be able to reach into other markets without it. In this clip from Streaming Media Connect 2025, execs from two larger operations, Google Cloud's Albert Lai and Vevo's Natasha Potashnik, share with IntelliVid's Steve Vonder Haar not only the cost savings opportunities but also how AI-enabled localization can drive new revenue opportunities as well.
24 Mar 2025
Is Google TV the smart TV platform of the future,or still just "beta" software?
04 Jan 2013
Companies and Suppliers Mentioned