Perplexity CEO Discusses Battling Google and AI Browser Showdown

Aravind Srinivas, the CEO of Perplexity AI, is reportedly challenging Google by aiming to have his company’s AI assistant preinstalled on Android devices. Simultaneously, according to an article by Alex Heath, Deputy Editor and author of the Command Line newsletter, Srinivas is positioning his startup to compete in what he believes will be the next major arena for AI competition: the web browser.

Srinivas disclosed that Perplexity is planning to launch its own browser, named Comet, in the coming month. He explained that “the rationale behind creating the browser is that it could be the optimal way to develop agents.” He elaborated, stating that “a browser essentially functions like a containerized operating system. It enables access to various third-party services via concealed tabs if you’re already logged in, allows client-side scraping, and can reason and act on your behalf.”

Heath’s article points out that other AI companies are also moving in this direction, citing OpenAI’s Operator and Google’s Mariner as examples that utilize the browser to execute commands and manage websites. While OpenAI has not yet released its own browser, there are ongoing rumors of its development. Meanwhile, Google might face a mandate from the US government to divest Chrome due to a ruling that found its search market dominance to be monopolistic.

The future of Chrome remains uncertain, but Heath notes that antitrust scrutiny on Google has created opportunities for Perplexity to secure distribution agreements with Android manufacturers. Motorola recently announced its decision to pre-install Perplexity on its new Razr phones, a move that could potentially bring Srinivas’s AI “answer engine” to millions more users. Srinivas views this partnership as a significant win for Perplexity, even if the level of integration is not as comprehensive as he would prefer.

According to Srinivas, “If Google hadn’t faced the DOJ trial, this partnership wouldn’t have materialized. They would have pressured many OEMs. I’ve had discussions with telcos who refused to engage with us, fearing revenue loss if Google found out.”

Heath recounts his previous conversation with Srinivas over a year prior, at which time Perplexity had around 1 million users and had raised under $100 million. In contrast, the article states that the startup now boasts nearly 30 million monthly active users and has raised hundreds of millions in funding. Perplexity is currently processing approximately 600 million queries monthly, which Heath notes is about 14% of Google’s query volume.

The interview between Heath and Srinivas took place the day before the Motorola announcement. During their conversation, they also discussed other potential partnerships aimed at expanding Perplexity’s reach, Srinivas’s strategy for establishing ownership of the browser interface, the development of an iOS assistant capable of controlling other apps, and his perspectives on managing TikTok, among other topics.

When Heath inquired about the genesis of the Motorola partnership and the challenges encountered with Google, Srinivas explained that discussions gained momentum after the demonstration of the Perplexity Android assistant, which launched in January. Motorola tested the assistant and found it to be significantly more reliable than Gemini. They were reportedly enthusiastic about preloading the app and prompting users to set Perplexity as the default assistant. However, Srinivas stated that Google intervened, indicating that they could not proceed with the phone’s launch on the Play Store and official Android version unless Gemini remained the default system. He emphasized that Google still maintains considerable control over the Android ecosystem, making it a process involving seven or eight clicks to change the default assistant.

Regarding Samsung, which has invested in Perplexity, Heath asked about the possibility of a similar partnership to the one with Motorola. Srinivas responded affirmatively, expressing his hope for collaboration. He noted that the specifics, such as who gets the default setting or whether it will be part of an onboarding process, were still under discussion.

Heath observed Srinivas’s apparent strong focus on distribution and partnerships as a means of growing Perplexity. Srinivas confirmed this strategy, stating their aim to collaborate with various entities, including ongoing work with telcos and plans to expand to OEMs. He also mentioned the upcoming launch of a browser with versions for Mac and Windows, with the intention of seeking OEM partnerships in that domain as well. Srinivas acknowledged the even more stringent contracts between Microsoft and OEMs on laptops compared to Google and Android, recognizing the uphill battle in achieving unbiased opinions favoring Copilot over Perplexity, despite Copilot being the natively loaded AI on Windows.

Concerning the recently launched iOS assistant, which surprised users with its capabilities, Heath questioned whether Apple had granted special permissions for controlling other apps. Srinivas clarified that Apple had not provided such permissions. He detailed the limitations, stating that their system cannot set alarms, enable low power mode, adjust brightness or volume, toggle the flashlight, make calls, or send iMessages. However, he explained that they leveraged the Apple EventKit SDK, which provides access to Reminders, Podcasts, Apple Music, Apple Maps, and other Apple apps. By utilizing Perplexity’s search infrastructure and deep linking to apps like YouTube and Uber, they address use cases where Siri falls short.

Heath then turned to the rationale behind developing a browser and its expected launch date. Srinivas reiterated that creating a browser is likely the optimal method for building agents, especially given the lack of OS-level control on iOS and Android. He described a browser as acting like a containerized operating system, facilitating access to third-party services, page scraping, reasoning, and action-taking on the user’s behalf, which he finds to be an appealing architecture. Srinivas stated that answering questions is becoming increasingly commoditized, and he believes Perplexity’s next advantage lies in performing actions, hence the development of the browser as the prime platform for user actions, prompting a shift to a new front-end.

Addressing the issue of publishers being unhappy with Perplexity’s content scraping, Heath noted the company’s move towards compensating some of them and asked if Srinivas felt they were now in a better position with publishers. Srinivas acknowledged that there is still work to be done but believes the situation has significantly improved since their previous conversation. He affirmed that Perplexity respects robots.txt and utilizes third-party data providers for non-scrapeable content.

Regarding reports of Perplexity raising hundreds of millions at an $18 billion valuation, Heath inquired about how this funding would be utilized. Srinivas explained that to build reliable agents, frontier reasoning models are essential. He anticipates current expenses dropping significantly within a year but emphasized the need to deploy to as many users as possible now to gather data, distill it into smaller models, and ultimately reduce costs.

Heath also asked about the status of Srinivas’s reported bid for TikTok and whether he had recently spoken to the White House. Srinivas stated that he had not abandoned the bid, but acknowledged that his chances were never the strongest. He clarified that funding was not the primary issue, as there were sufficient backers. Instead, he believes ByteDance’s reluctance stems from their desire to maintain control over the algorithm, as they believe no one else can manage it as effectively. Srinivas also noted the deep intertwining of the app’s operations in America and Europe, making separation difficult, and mentioned that tariffs will influence everything, including TikTok.

Finally, Heath questioned whether Srinivas foresaw ChatGPT’s scale impacting Perplexity’s user base, particularly given ChatGPT’s potential for personalization creating user lock-in. Srinivas responded by outlining ChatGPT’s strategy of integrating a “Login with ChatGPT” button on third-party apps to gather data, but noted that this requires convincing app developers to implement it. He contrasted this with Perplexity’s approach, which allows users to stay logged in where they are. He believes that by building a browser, Perplexity can access apps on the client side on behalf of the user.

Srinivas posited that memory will be dominated by the company with the most context. He argued that ChatGPT lacks knowledge of user interactions on platforms like Instagram or Amazon, as well as time spent on various websites. He believes that deep personalization requires comprehensive data, not just retrieval of past queries. He sees the challenge as integrating transactions, commerce history, and browser activity into the assistant across platforms, which necessitates developing browsers for both web and mobile and sharing cookies across apps.

When Heath suggested that Srinivas sees the browser as the ultimate goal for Perplexity’s development, Srinivas clarified that there is more beyond that, including building solutions for Windows, Mac, Android, and iOS. He views a browser as limited and containerized, with the operating system representing the final frontier.