How AI Upended a Historic Antitrust Case Against Google

When the United States Justice Department first sued to break up Google alleging that it illegally monopolized online search in October 2020, there was little indication that one of the biggest factors in the case would be the rapid rise of a nascent technology.

On Tuesday, US District Court Judge Amit P. Mehta ordered Google to stop using exclusive agreements with third-parties to distribute its search engine, but stopped short of forcing the company to cease such payments altogether or to spin off its Chrome web browser.

The decision over legal remedies in the case deals a significant blow to US antitrust enforcers after securing a historic ruling declaring that Google maintained an illegal monopoly last year.

Notably, Mehta’s 226-page liability decision heavily emphasized the role that the ascendance of artificial intelligence, particularly generative AI (or “GenAI”) products like OpenAI’s ChatGPT, played in his assessment of the case.

“The emergence of GenAI changed the course of this case,” Mehta wrote in his 226-page ruling.

Tech Policy Press reviewed Mehta’s mentions of AI tools and companies and his characterization of Google’s position in this emerging market to see how his assessment of the technology impacted his deliberations. Here’s what we found:

A blip during liability discussion, a major talking point over remedies

Google’s competitive position in the booming yet still emerging AI market featured prominently in Mehta’s decision Tuesday, a contrast with his earlier ruling finding that Google monopolized online search. As CNBC reported, “OpenAI’s name comes up 30 times, Anthropic is named six times and Perplexity shows up in 24 instances. …ChatGPT was named 28 times in Tuesday’s filing.” Aside from OpenAI, the companies had not yet been founded when the case was filed.

Additionally, “AI” and “artificial intelligence” were mentioned 116 times combined, generative “artificial intelligence,” “generative AI” and “GenAI” were referenced 220 times, and “large language models” and “LLM” were mentioned 82 times, according to our review.

By contrast, Mehta barely made reference to AI’s rise in his decision declaring Google a monopoly last year. In that 286-page decision, Mehta mentioned ChatGPT only twice, and OpenAI, Perplexity and Anthropic not at all. “Generative artificial intelligence” was mentioned seven times, while “generative AI” and “GenAI” were not referenced at all, and “large language model” and “LLM” were referenced only a dozen times.

Mehta himself alluded to this discrepancy, noting that the tools played a far bigger role in the latter remedies phase of the trial than the earlier liability phase. While no AI competitors have yet to make gains on Google, Mehta wrote, the tools “may yet prove to be game changers.”

No witness at the liability trial testified that GenAI products posed a near-term threat to GSEs [general search engines]. The very first witness at the remedies hearing, by contrast, placed GenAI front and center as a nascent competitive threat. These remedies proceedings thus have been as much about promoting competition among GSEs as ensuring that Google’s dominance in search does not carry over into the GenAI space.

Projecting AI’s path in search

Mehta lamented that the case required the court to “gaze into a crystal ball and look to the future,” which he said was not “exactly a judge’s forte.” But he sought to do just that and paint a picture of how AI tools are now and could soon intersect with Google’s grip over search.

Mehta wrote that “tens of millions of people use GenAI chatbots, like ChatGPT, Perplexity, and Claude, to gather information that they previously sought through internet search,” and that experts expect generative AI tools to increasingly perform like search engines.

“Like a GSE, consumers can interact with AI chatbots by entering information seeking queries. … Thus, chatbots perform an information-retrieval function like that performed by GSEs,” he wrote, though he noted chatbots can also perform distinct functions, like generating images.

Their aim, he wrote, is to “transform chatbots into a kind of “[s]uper [a]ssistant” able to perform “‘any task’” asked by a user. “Search is a necessary component of this product vision,” he concluded.

Mehta also considered current evidence that the tools are already factoring into the online search landscape. While he noted that Google may now be using its own AI tools to strengthen its dominance over search, a key concern for US authorities, he also wrote that “GenAI products may be having some impact on GSE usage,” and that competitors are also looking to use AI tools to onboard users onto their products as “access points” for search queries. Mehta alludes to the vision shared by AI firms that one such access point may eventually be a “super assistant” that “would be able to help perform ‘any task’ requested by the user.”

A “highly competitive” AI market

In his discussion of the current generative AI market, Mehta described it as “highly competitive” with “numerous new market entrants” in recent years, including the Chinese firm DeepSeek and Elon Musk’s Grok, and wrote that Google is not exactly in pole position to dominate it.

“There is constant jockeying for a lead in quality among GenAI products and models … Today, Google’s models do not have a distinct advantage over others in factuality or other technical benchmarks.”

He listed Anthropic, Meta, Microsoft, OpenAI, Perplexity, xAI, and DuckDuckGo as other participants in the market, and noted that they “have access to a lot of capital” to compete.

Mehta also wrote that generative AI companies have “had some success” in striking their own distribution agreements with device manufacturers to place their products, including partnerships between OpenAI and Microsoft and Perplexity with Motorola.

This section echoed many of the points Google made in its defense. Last year, the company wrote in a blog post about the case that the court was evaluating a “highly dynamic” market. “Since the trial ended over a year ago, AI has already rapidly reshaped the industry, with new entrants and new ways of finding information, making it even more competitive,” Google wrote.

The company has said it plans to appeal the initial liability ruling finding that it maintained an illegal monopoly, while in a statement released following the decision DOJ leaders appeared to suggest they may appeal the remedies Mehta doled out this week.

Some solace for US enforcers

While Mehta’s decision was far less sweeping than US antitrust enforcers had hoped for, his remedies will impact Google’s relationship with its budding AI rivals.

Mehta ordered Google to cease exclusive distribution agreements and share some of the data it uses to power its search business, including with companies in the AI space.

Because their functionality only partially overlaps, GenAI chatbots have not eliminated the need for GSEs. … Nevertheless, the capacity “to fulfill a broad array of informational needs” constitutes a defining feature of both products, as Google implicitly acknowledges. … And it is that capacity that renders GenAI a potential threat to Google’s dominance in the market for general search services.

But Google’s seeming inability to significantly leverage its dominance in search to quickly boost its AI offerings appeared to be a major sticking point for Mehta in weighing tougher sanctions.

The evidence did not show, for instance, that Google’s GenAI product responses are superior to other GenAI offerings due to Google’s access to more user-interaction data. If anything, the evidence established otherwise: The GenAI product space is highly competitive, and Google’s Gemini app, for instance, does not have a distinct advantage over chatbots in factuality and other technical benchmarks.

Mehta did leave the door open that if the situation changes, the court could intervene more substantially. Market “realities give the court hope that Google will not simply outbid competitors for distribution if superior products emerge,” Mehta wrote. “The court is thus prepared to revisit a payment ban (or a lesser remedy) if competition is not substantially restored through the remedies the court does impose.” Presumably that determination would be informed by the work of the Technical Committee established by the court, which is set to function throughout the six-year term of the judgment.