Join Puck to listen to this article
Last year at Google I/O, the company’s annual developer conference, C.E.O. Sundar Pichai stood before the crowd and heralded a “total reimagining of Search” with the introduction of Google’s A.I. Mode. This year, he declared that Search is “bringing the benefits of generative A.I. to more people than any product in the world.” He called A.I. Mode a “revelation,” described the impending A.I.-powered “transformation” of Search as “our ultimate moonshot,” and announced that we’ve entered the company’s “agentic era.”
It wasn’t total hyperbole. The broad rollout of A.I. Mode will introduce a new user interface—an “intelligent Search box,” powered by Gemini’s 3.5 Flash model—that conveys information in an interactive, personalized format, with links to YouTube videos (and ads) running alongside and the ability to ask follow-up questions. “We’re entering the next chapter of Google Search, where incredible A.I. features aren’t just in Search—Google Search is A.I. search, through and through,” Google Search head Liz Reid said during the keynote.
Of course, the new function—which was introduced last year as an experimental feature—produced plenty of anxiety among publishers, who are already navigating a precipitous decline in search traffic as a result of Google’s A.I. Overviews. (My partner Julia Alexander wrote brilliantly about this problem last week.) But as A.I. Mode is introduced to the entirety of Google’s users, another, adjacent concern has emerged: what the feature means for the quality of information that appears at the top of the search page.
Ranjit Singh, the director of Data & Society’s A.I. on the Ground program, explained that there’s a big difference between users self-selecting links on Google and an A.I. system (with unclear decision-making processes) sorting through those links and summarizing them for you. In short, he said, it can be unclear why information from certain websites would be chosen for summarization while other sources are ignored—a problem that experts in algorithmic bias have frequently raised with me. This is not necessarily a new problem, obviously; traditional search ranking algorithms do something similar, and Google has a whole list of signals that its algorithms check to determine site rankings. But the jump to information generation takes that a step further. (A Google spokesperson told me that A.I. Mode is not the default Search experience, and added that its A.I. experiences are designed to show links prominently throughout and beside outputs.)
One potential result of this change is that the additional layer of mediation could risk complicating the reliability of the company’s core product. Indeed, Google is “no longer in the space of just ranking different sources of information,” Singh said. “They’re in the space of generating that information—an important role, and a very different role for Google.” At the same time, he noted, the company’s evolution into a quasi content creator was inevitable. “In the last couple of years, the way in which we search for information has changed,” he said. “A lot of people just ask chatbots for answers.”
Concerningly for Google’s hyperscaler competitors, users seem to be largely embracing its A.I. features: According to the company, A.I. Mode has racked up a billion monthly users since its rollout last year, A.I. Overviews fields about 2.5 billion monthly users, and the Gemini app has just shy of a billion. So maybe the company has occupied this new space after all.
“Google’s Game to Lose”
Of course, undermining the internet’s gazillion publishers comes with certain risks. In theory, disincentivizing websites from producing content should put a real dent in Google’s ad revenue, to say nothing of the quality of information that A.I. Mode is able to source. But Nikhil Lai, a principal analyst at Forrester, noted that Google doesn’t make that much (relatively speaking) from sharing ad revenue with publishers, which accounts for only a small portion of the tens of billions it earns from search ads.
Lai added that the only real risk for Google is if advertisers start throwing more money at S.E.O. practices to nestle themselves into A.I. outputs than they do at advertising. “One client of ours spends $1.2 million on S.E.O., and like, $220 million on paid search. If that dynamic changes, that’s a problem,” he said. Malik Ahmed Khan, a senior analyst for Morningstar, added that Google has also been working for a long time to integrate features intended to keep users on its own website longer—which means, yes, the ability to serve more ads. “They wouldn’t be making this transition unless they were confident that they had nailed the monetization angle, because it’s way too high stakes for them,” he said.
Then there’s the fact that A.I. Mode doesn’t require a subscription, unlike ChatGPT, Claude, and other A.I. competitors. “There is actually no other mechanism to bring these chatbots to consumers at scale without advertising,” Khan said, noting that Google already has that industry-defining advertising infrastructure firmly in place. He told me this vastly expands Google’s monetization potential for A.I. Mode, even if the overhaul of Search might marginally impact revenue elsewhere.
Indeed, Google’s competitors seem to already acknowledge the company’s edge with consumers. “I think the competition is kind of clearing the path for them,” Khan said, pointing out that both Anthropic and OpenAI have started to focus on enterprise customers over the kind of costly, consumer-oriented A.I. experience that’s becoming Google’s forte. And if this reimagination of Search is successful, “the reason why anyone would actually go and use ChatGPT just goes away completely in my opinion,” he told me. Lai, for his part, put it more bluntly: “I think this is Google’s game to lose.”
The Great Differentiation
Still, ever since A.I. Overviews’ rocky start a few years ago, there’s been an impression that Google might be alienating users, many of whom want the option to toggle A.I. features on and off. But that hasn’t panned out: Google’s share of the search market is still around 90 percent, exactly where it’s hovered for years. Nonetheless, it’s created an opening for Search competitors that do things differently. “I think that is an opportunity for us, especially combined with the invasion of privacy,” said Kamyl Bazbaz, the chief communications officer for DuckDuckGo. “I do think we’re in a pre-Snowden, pre-Cambridge Analytica moment for the A.I. era, but it’s going to happen, and when that happens, I think people will turn to us.”
Bazbaz told me that the platform, which offers private search browsing and A.I. search options that users can toggle on and off, has experienced “double-digit percentages in growth every year” since ChatGPT came out. “A.I. companies aren’t listening to user preference, and we’re the complete opposite. We have to listen to it. Our future and our existence kind of depend on it,” he told me. “So we have to design A.I. for people that hate A.I. and just want to turn it off, and we’re just sitting right at this tension, and hope we can ride it out and grow, really, because people want those options.”
Lai, however, thinks the average user’s impulse to query Google is just too entrenched for that kind of differentiation to really have an impact. Ironically, he said, distrust in A.I. is actually strengthening Google’s position—even as the company expands its own A.I. offerings. “Google is a more credible purveyor of information, where we’re seeing many consumers, once they suspect that they’re being lied to by ChatGPT, going back to Google to verify information they’re suspicious of,” he said. “I’m not seeing any shift toward [companies like Brave or DuckDuckGo] because of distrust in A.I.”
The question now is whether any of the other hyperscalers can figure out how to maneuver around Google’s incumbency advantage—because right now, it looks like no one can really compete with them.
