The Verdict Everyone Wanted, and the Problem Nobody Solved
In August 2024, a federal judge handed the United States Department of Justice one of the most consequential antitrust victories in a generation. Judge Amit Mehta ruled that Google had illegally maintained a monopoly in the general search market, finding that the company had spent billions of dollars each year paying Apple, Samsung, and other device manufacturers to make Google the default search engine on their products. The ruling was unambiguous: Google had broken the law.
And yet, two years later, the question of what to do about it remains genuinely unresolved. After the remedies phase, Judge Mehta ruled in September 2025 that Google would not have to sell its Chrome browser, but he barred exclusive default deals and ordered the company to share certain search data with qualified competitors. The case is now on appeal, with Google challenging the underlying ruling and the government asking the appeals court to revisit the remedies. Meanwhile, the search market itself has been quietly transformed by a technology that none of the litigants anticipated when the DOJ filed its complaint in 2020: generative artificial intelligence.
The story of antitrust law and search engines is a story about a legal system built for the industrial age trying to regulate a product that behaves unlike anything that came before it. It is also a story about how winning a case and actually fixing a market are two very different things.
How Google Built a Wall Others Cannot Climb
To understand what antitrust law is working with, it helps to understand precisely how Google’s dominance was constructed. The company controls close to 90 percent of global general search queries, a figure that has remained remarkably stable for years despite the existence of well-funded competitors including Microsoft’s Bing, DuckDuckGo, and others.
Judge Mehta’s ruling identified the default placement deals as the central mechanism of that dominance. Google reportedly paid Apple about 18 billion dollars in 2021 alone, part of the 26.3 billion dollars it paid distributors that year according to figures disclosed during the trial, to remain the default search engine in Safari across iPhones, iPads, and Mac computers. Similar agreements covered Android devices from Samsung and other manufacturers. The effect was to capture users at the precise moment they were most likely to form habits, before they had any reason to try something else.
This matters because of what economists call the “query feedback loop.” Search engines improve by processing enormous volumes of queries. Each search tells the engine something: which results the user clicked, how long they stayed, whether they came back to search again. Google processes billions of searches per day. Bing processes far fewer. The gap in data volume translates directly into a gap in result quality, which translates into user preference, which generates more data. The wall, once built, becomes self-reinforcing.
The DOJ argued, and Mehta agreed, that Google’s payments to distributors prevented rivals from ever accumulating the query volume necessary to close that gap. Competing on the merits, in other words, was structurally impossible when the dominant player had locked up the most important distribution channels.
What the Law Can Actually Compel
Antitrust remedies in the United States flow from Section 2 of the Sherman Antitrust Act, a piece of legislation passed in 1890, long before the internet existed. The law prohibits monopolization but leaves the question of remedy largely to judicial discretion. Courts have historically chosen between two broad categories: behavioral remedies, which order a company to stop certain practices, and structural remedies, which physically break a company apart.
The most famous structural remedy in American history was the breakup of AT&T into seven regional “Baby Bells,” settled in 1982 and carried out in 1984. The DOJ famously sought to break up Microsoft in 2000, though that remedy was eventually rejected on appeal and replaced with behavioral restrictions. Structural remedies are dramatic, but courts have grown increasingly reluctant to impose them absent compelling evidence that behavior alone cannot restore competition.
In the Google case, the DOJ’s proposed remedies evolved considerably. Prosecutors pushed for prohibitions on exclusive default agreements, requirements that Google share certain search data with competitors, restrictions on how Google uses its browser and operating system to favor its own search product, and the forced divestiture of Chrome. Mehta declined to order the divestiture of Chrome or Android, but he banned exclusive distribution contracts and ordered Google to share certain search data with qualified competitors, while still permitting non-exclusive payments for default placement. Google’s lawyers have argued that sharing proprietary data would harm user privacy, and stifle innovation.
Legal scholars are divided on how far a court should go. Behavioral remedies are easier to justify legally, but they may not be sufficient, because a remedy has to actually change competitive conditions and not just stop the specific conduct that was found to be illegal. The concern is that even without the default agreements, Google’s accumulated data advantages and brand recognition may be enough to sustain its dominance indefinitely.
There is also the practical problem of enforcement. Behavioral remedies require ongoing monitoring. The Microsoft settlement of 2001 spawned years of compliance disputes. A court can order Google to stop exclusive default deals, but ensuring the company does not find subtler ways to achieve the same end requires sustained regulatory attention that American institutions have historically struggled to provide.
The AI Disruption: A Gift or a Distraction?
Hovering over all of this is the question that has dominated technology commentary since late 2022: will AI-powered search simply render the Google monopoly problem moot?
The theory of disruption goes like this. ChatGPT, Perplexity, Microsoft’s Copilot integration into Bing, and Google’s own Gemini-powered AI Overviews represent a fundamental shift in how people find information. If users increasingly get answers from conversational AI systems rather than ten blue links, the traditional search market may fragment on its own, without regulatory intervention.
There is some evidence for this view. Perplexity AI has grown quickly as an AI-native search product. Microsoft has integrated AI capabilities throughout its Edge browser and Bing product, hoping to convert its second-place position in traditional search into a first-mover advantage in AI search. OpenAI launched its own search product in 2024, bringing one of the most widely used AI systems directly into competition with Google.
But the disruption thesis has significant holes. First, Google was not caught flat-footed. The company launched AI Overviews broadly in 2024 and has invested heavily in Gemini across its products. Its distribution advantages, the same ones that made its traditional search monopoly so durable, apply equally to AI search. Google Assistant and Gemini are preloaded on many Android devices. Distribution deals like these can extend to AI features as well. The query feedback loop may be even more important in AI search than in traditional search, because training large language models requires massive amounts of data.
Second, AI search has introduced a new set of competitive concerns that antitrust frameworks have not yet addressed. When an AI system synthesizes an answer from multiple web sources, it often does not send users to those sources, potentially destroying the economic model of the publishers whose content trained the system. Perplexity and Google have both faced lawsuits and complaints from news organizations over this issue. The question of who owns the data that makes AI search possible is becoming its own legal battleground, separate from but entangled with the antitrust proceedings.
Third, and perhaps most importantly, the concentration of AI infrastructure may be replicating, or even amplifying, the dominance of the existing tech giants. Training frontier AI models requires extraordinary amounts of compute, data, and capital. The companies best positioned to win the AI search race are, largely, the same companies that already dominate the digital economy: Google, Microsoft, and Amazon. A new technology does not automatically produce a new competitive order.
Europe’s Different Approach and Its Limits
The United States is not the only jurisdiction grappling with search dominance, and comparing approaches reveals both the possibilities and the limitations of different legal frameworks.
The European Commission fined Google 2.42 billion euros in 2017 for favoring its own shopping service in search results, 4.34 billion euros in 2018 over Android, and 1.49 billion euros in 2019 over search advertising restrictions. These are large numbers, but critics noted that they amounted to a fraction of the company’s annual profits, functioning more as a cost of doing business than a genuine deterrent.
More recently, Europe’s Digital Markets Act, which came into force in 2022 and began applying to “gatekeepers” in 2024, represents a fundamentally different regulatory philosophy. Rather than waiting for harm to occur and then prosecuting it, the DMA imposes ex-ante obligations on designated platforms. Under the DMA, Google is required to allow users to change their default search engine easily, to give rivals fair access to certain data, and to refrain from self-preferencing its own products in results.
The early results have been mixed. Default search choice screens, mandated by European regulators, have produced modest shifts in market share in some countries but have not fundamentally disrupted Google’s position. Users, when given a choice, often simply select Google anyway, a finding that cuts both ways: it suggests genuine consumer preference, but it also reflects years of habituation and brand familiarity that the choice screen alone cannot undo.
Some competition lawyers argue that the DMA’s real value may not be in its early enforcement actions but in its deterrent effect and its data-sharing provisions, which could, over time, allow rivals to accumulate the query volume needed to compete meaningfully. Others argue that the DMA’s complexity and the difficulty of defining terms like “fair access” will bog it down in years of definitional litigation, much as American behavioral remedies have historically done.
What Antitrust Cannot Fix
There is an important and underappreciated category of things that antitrust law, regardless of jurisdiction, is structurally ill-equipped to address when it comes to search.
Antitrust law is fundamentally about process: it asks whether competition was distorted by illegal conduct. It is not, in its traditional form, a tool for engineering particular market outcomes. A court can order Google to stop exclusive default deals. It cannot order consumers to use Bing. A court can require data sharing. It cannot guarantee that a competitor will use that data to build a genuinely better product, or that users will discover it if they do.
There is also the “zero price” problem. Search engines are free to consumers. Traditional antitrust analysis focuses heavily on price harm, asking whether a monopolist has raised prices above competitive levels. When the product is free, that framework requires adaptation. The DOJ and academic economists have argued that harm should be measured in reduced quality, less privacy, and advertising market distortions rather than consumer prices. That framing featured in the Google case, but it remains contested and complicates the remedy design.
Perhaps most fundamentally, antitrust cannot resolve the underlying policy question of what we actually want from a search market. Do we want many small search engines of roughly equal quality? A regulated utility model, where one dominant engine operates under public service obligations? Or a vibrant AI search market where new entrants compete on the quality of their synthesis and reasoning, even if the underlying data infrastructure remains concentrated? These are political and social questions that antitrust law, on its own, has no mechanism to answer.
The Long Game: What Comes After the Ruling
With the Google case now in the appeals phase, a few things seem relatively clear and several remain genuinely uncertain.
The court has already rejected the structural remedy of divesting Chrome, and the ban on exclusive default agreements and the data-sharing requirements are the core of the order. Both sides have taken the case to the D.C. Circuit, and the government has asked the court to reconsider the rejected ban on payments for default placement, meaning the effective implementation of the order may be years away.
The AI dimension will complicate enforcement regardless of what the court orders. Defining the relevant market for purposes of a remedy, a market that looked like “general search” in 2020, now encompasses AI assistants, large language model search interfaces, and vertical AI tools, will be contested at every turn.
What the case has already accomplished, regardless of its eventual outcome, is a shift in the political economy of tech regulation. The ruling established, as a matter of American law, that a technology company can illegally maintain a monopoly through distribution agreements rather than superior product quality alone. That precedent applies beyond Google and beyond search. It is available to future prosecutors and plaintiffs challenging dominance in AI, cloud computing, or whatever market emerges next.
The deeper lesson may be this: antitrust law is a necessary but not sufficient tool for addressing digital market dominance. It can remove some of the scaffolding that holds a monopoly in place. It can open space for rivals to compete. But it cannot substitute for affirmative industrial policy, data governance frameworks, or the kind of sustained regulatory capacity that Europe has tried, imperfectly, to build through the DMA. The verdict against Google was a beginning, not an end. The harder work of actually making search markets competitive is still very much underway.