Browser Wars: The Rise of AI

Sign up for The Media Today, CJR’s daily newsletter.

Last year, after a US district court judge found that Google had illegally monopolized the market for online search, the Department of Justice and state attorneys general proposed a range of remedies, including requiring Google to divest from its Web browser, Chrome. With over 66 percent of the global browser market, Chrome far outpaces competitors like Apple’s Safari (which has around 18 percent of the market) and Microsoft’s Edge (5 percent); in turn, this dominance reinforces the supremacy of its default search engine, Google Search, which commands nearly 90 percent of the worldwide search market. (Mathew Ingram wrote about the various antitrust cases facing Google and other tech titans in this newsletter last week.)

This week, during the second week of the remedy phase of the trial, the Department of Justice said that Google’s search monopoly gives it an unfair advantage in artificial intelligence development. Prosecutors argued that the company’s ability to leverage its vast user data repository and extensive search index to train its AI tools reinforces its dominance and creates barriers to competition. As such, the government is pursuing structural remedies that could prevent Google from consolidating power in the emerging AI sector, including potential mandates to share search data with competitors. Google, for its part, argues that the AI market is already highly competitive and that its rivals are thriving without access to its proprietary data.

Regulatory pressure is only one of the forces challenging Google’s long-standing dominance in this area—the rapid rise of generative AI has sparked renewed innovation and competition in the browser and search markets, which had long been seen as stagnant. Browser developers are racing to integrate AI-driven features aimed at improving personalization, efficiency, and user engagement, including personalized search results, smart content recommendations, real-time translation, automated summaries, and enhanced security features. Microsoft’s Edge now features Copilot, the company’s AI assistant; the Brave browser released an AI assistant called Leo; Firefox recently added support for third-party chatbots; and Apple has integrated Apple Intelligence summaries into Safari. Opera, which launched its Aria AI assistant in 2023, recently previewed a new “agentic” tool called Browser Operator, which can independently perform browsing tasks on users’ behalf; the Browser Company is developing its own “agentic” browser called Dia. 

AI companies, meanwhile, are eyeing the browser market. In November, shortly after integrating real-time Web search into its chatbot, OpenAI was reported to be looking into developing a browser that integrates its chatbot and search capabilities, potentially in partnership with e-commerce platforms; as Inc.’s Kit Eaton noted, such a move would help the company retain users within its ecosystem and open new monetization avenues via subscriptions or ads, all while placing ChatGPT at the centers of users’ online experiences. In February, Perplexity announced plans to launch its own browser, Comet, which is set for release next month. On the TBPN podcast last week, Aravind Srinivas, Perplexity’s CEO, explained that one of the motivations behind building a browser is the ability to collect richer behavioral data to support ad-targeting: “We plan to use all the context to build a better user profile, and maybe, you know, through our ‘discover’ feed we could show some ads there.”

These companies aren’t exclusively eyeing the development route: this week, OpenAI, Perplexity, and Yahoo have all expressed interest in acquiring Chrome, should Google be forced to divest it. (DuckDuckGo’s CEO said that that company wouldn’t be able to afford it.) A browser with Chrome’s reach would be immensely valuable for any company aiming to scale its search operations. As Brian Provost, the general manager of Yahoo Search, testified this week, approximately 60 percent of user queries originate from a browser’s search bar. Buying Chrome would not only make an AI company significantly more visible, but would also give it a staggering amount of data to train its models, Futurism’s Noor Al-Sibai wrote.

Yet the rush to embed AI into browsers and search tools seems to ignore serious concerns about the impact of this technology on information quality. At least one browser, Vivaldi, is holding out on the trend, explicitly rejecting AI integration due to concerns about accuracy, ethics, and environmental impact. Large language models “are essentially confident-sounding lying machines with a penchant to occasionally disclose private data or plagiarise existing work,” Julien Picalausa, a software developer at Vivaldi, wrote in a blog post. “As such, it does not feel right to bundle any such solution into Vivaldi.” In research I have conducted with my Tow Center colleague Aisvarya Chandrasekar, we’ve found that generative search tools often fail to reliably cite their sources and frequently deliver inaccurate information with unwarranted confidence. There are countless instances of chatbots producing questionable outputs; last week, social media users shared amusing examples of Google’s AI Overviews confidently explaining fake idioms like “you can’t lick a badger twice.” Worse still, these systems remain vulnerable to manipulation. An investigation published in March by NewsGuard found that major AI chatbots repeated false claims promulgated by a pro-Kremlin disinformation campaign. 

Generative AI tools also appear to be cannibalizing traditional Web traffic. A recent study by Ahrefs, a provider of SEO tools, found that Google’s AI Overviews reduce click-through rates to top-ranked websites by more than 34 percent. The growing popularity of search summarization tools could have severe economic consequences for publishers, including news companies, that depend on search traffic for advertising revenue. (Earlier this year, the education technology company Chegg sued Google, claiming that its AI-generated summaries of search results had substantially reduced its site traffic and revenue.) Publishers have also raised alarm about the significant increase in the volume of requests being driven by bots that “scrape” websites to collect data that goes into training generative AI models. Last month, Wikimedia published a post highlighting how automated requests for its content have grown exponentially, without sufficient attribution, causing “a significant load on the underlying infrastructure that keeps our sites available for everyone.”

Some observers, like Scott Rosenberg of Axios, are drawing parallels between this moment and the “browser wars” that took place in the nineties between Microsoft and Netscape. Although Microsoft won that initial battle with its Internet Explorer browser—prompting a landmark antitrust lawsuit—it eventually lost ground to more innovative challengers like Mozilla and Chrome. “Tech’s platform shifts may feel high-velocity,” Rosenberg wrote, “but they take a long time to unfold—and you should never feel too sure you know who is going to own the future.” Change is surely afoot, though. Alissa Cooper, the executive director of a tech research and policy center at Georgetown University, told NPR this week that, no matter the outcome of the Google remedy trial, “it will be having some kind of impact on AI.”

Other notable stories:

All this week, CJR is running a series of pieces, on our website and in this newsletter, on the fog of news and propaganda that has marked the first hundred days of Donald Trump’s second term as president. This morning, Jack Herrera reports on how journalists have struggled to cover Trump’s push to offshore his migrant detention regime. “Before this year, reporters already struggled to cover the vast, hard-to-access archipelago of immigrant detention centers across the US. That chain of jails now stretches deep into Latin America,” Herrera writes. Since Trump began speeding up the deportation of migrants alleged to be gang members without due process under an old wartime law, the time between arrest and deportation has sometimes been “just hours,” Herrera adds, “which makes press access to these deportees a near impossibility.”

As we noted in yesterday’s newsletter, lawyers for Trump and Paramount entered mediation yesterday in a bid to resolve a lawsuit that Trump filed against CBS, which Paramount owns, last year; according to the Wall Street Journal, however, the two sides appear to be far apart, with Trump wanting more money than Paramount initially seems willing to offer, as well as an apology. On Truth Social yesterday, Trump accused CBS of perpetrating “one of the most egregious illegalities in Broadcast History,” and also hinted that he might sue the New York Times for (accurately) reporting that many legal experts view Trump’s case as baseless. Meanwhile, Puck reports that Paramount’s owner asked CBS to delay sensitive reporting on Trump while a corporate merger goes through.

Over the weekend, a huge explosion rocked a port in Iran, killing dozens of people; then, on Tuesday, an explosion at a munitions factory in the country killed at least one person. Iranian officials have blamed the port explosion on improperly stored cargo, rather than foreign sabotage, but according to The Guardian’s Patrick Wintour and Deepa Parent, journalists who have attempted to investigate the incidents have faced repression. “Suspicion of cover-ups is rife,” Wintour and Parent wrote on Tuesday, “and the filing of criminal charges against media outlets and activists by the Tehran prosecutor’s office has caused journalists in Iran to voice concern about press freedom.”

In March, Joakim Medin, a Swedish journalist working for a newspaper in Denmark, flew to Turkey to cover unfolding mass protests against the regime of Recep Tayyip Erdoğan, the increasingly authoritarian president, only to be detained at the airport and charged with insulting Erdoğan and belonging to a terrorist group (seemingly over photos that editors picked to accompany articles Medin wrote). Yesterday, a court handed Medin a suspended sentence on the first charge; the court ordered his release, but he will stay behind bars for now, pending resolution of the second charge. Agence France-Presse has more.

And the White House launched a new feature on its website, “WHWire,” that highlights positive stories about the administration and looks uncannily like the Drudge Report. The latter site soon mocked the former, splashing headlines like “IT TAKES AN ENTIRE WEST WING TO COMPETE WITH DRUDGE” and “TRUMP LAUNCHING HIS OWN DRUDGE REPORT…USING TAXPAYER DOLLARS!” Matt Drudge, the eponymous founder, who has soured on Trump, quipped that he is “considering a $1 trillion lawsuit!”