When the Department of Justice sued Google over search distribution in October 2020, ChatGPT did not exist. When the Federal Trade Commission sued Facebook two months later, TikTok had been available in the United States for roughly two years, and short-form video was just a feature, not the organizing principle, of social media. Five years on, two federal judges issued decisions illustrating the same economic fact: tech innovation is accelerating, challenging incumbents, driving innovation, and benefiting consumers.
Judge Amit Mehta and Chief Judge James Boasberg reached that same conclusion from different starting points and with different results for the government. In the Google case, dynamism shaped the remedy after a liability finding. In the Meta case, dynamism defeated the government’s market definition outright. The cases may have been distinct, but each applied a consistent principle; a principle that is highly relevant as Congress weighs considering AICOA again more than five years after it was first introduced.
Boasberg Finds Market Definition Is Date-Dependent
The FTC’s theory against Meta depended on a market for “personal social networking services” consisting of Facebook, Instagram, Snapchat, and the smaller MeWe. TikTok and YouTube were excluded on the theory that they exist for entertainment while Meta’s apps exist for keeping up with friends and family. That framing may have described 2012. Judge Boasberg’s November 18, 2025 opinion found that it did not describe 2025, and that 2025 was the relevant year.
Because the FTC sought a permanent injunction under Section 13(b) of the FTC Act, Boasberg held that the agency had to prove an ongoing or imminent violation, not merely past misconduct. The court held that the FTC had to prove Meta holds monopoly power now, and where the data allowed, it calculated market shares using the latest available 2025 figures. The evidence on how people actually use these products was lopsided. January 2025 data showed U.S. Facebook users spending only about 17 percent of their time viewing content from friends, and Instagram users about 7 percent. A majority of time on each app was spent watching video. Judge Boasberg found that Facebook, Instagram, TikTok, and YouTube had evolved to have nearly identical main features: users spend most of their time watching videos, algorithms pick those videos, and every app lets users forward what they like to friends.
The court required the FTC to show that Meta still holds monopoly power today, regardless of whether it had in the past. Meta’s paid-switching experiment, which measured where users actually went when compensated to spend less time on Meta’s apps, was described as the single best evidence of what consumers treat as substitutes. Outage data reinforced that conclusion, while the survey evidence failed to establish the sharp boundary the FTC proposed. Judge Boasberg’s summary was blunt: YouTube and TikTok belong in the relevant product market, refuting the FTC’s assertion that Meta still enjoys monopoly power today. The court further noted that even if YouTube were excluded, including TikTok alone in the relevant market defeats the FTC’s case.
What makes the opinion a landmark for dynamic-market analysis is how explicitly the court confronted the passage of time. The opinion opens with Heraclitus and the observation that the landscape of five years earlier, when the FTC brought suit, had changed markedly, and that the wall once separating social networking from social media had broken down. A market definition constructed around user behavior at the start of an investigation can become a historical artifact by the time trial ends. The court declined to treat that artifact as evidence of present power.
Judge Mehta Finds Remedies Must Fit the Market Today
Judge Mehta’s path was different. His noteworthy opinion covered the remedies phase in antitrust litigation against Google regarding general search and general search text advertising. The remedies question was what to do if Google was liable, and the answer turned on what had happened to the market in the interim.
The remedies opinion issued September 2, 2025 states the matter plainly: “The emergence of GenAI changed the course of this case. No witness at the liability trial testified that GenAI products posed a near-term threat to GSEs. The very first witness at the remedies hearing, by contrast, placed GenAI front and center as a nascent competitive threat.” The court found that companies building large language models are better positioned, both financially and technologically, to compete with Google than any traditional search company has been in decades, with the possible exception of Microsoft. Those developments materially influenced the court’s approach to relief. Alongside concerns about causal fit and downstream harm, emerging AI competition counseled against some of the government’s broadest proposals. Judge Mehta rejected Chrome divestiture and a wholesale ban on default-placement payments, while imposing restrictions on exclusivity and requiring data and syndication access.
The remedies opinion is direct about its reasoning. Courts, Judge Mehta wrote, must approach the task of crafting remedies with a healthy dose of humility. The court acknowledged that it has no expertise in the business of general search engines, the buying and selling of search text ads, or the engineering of generative AI. That is the correct posture for a generalist court asked to restructure a sector in which the competitive frontier is moving faster than the litigation calendar. A remedy calibrated to the market of 2020 would have been aimed at a target that no longer sat where the plaintiffs had drawn it. Judge Mehta’s approach was not to replace yesterday’s frozen picture with today’s. He left room to revisit a payment ban if the adopted remedies failed to substantially deliver intended outcomes.
Different Cases Reached The Same Principle
The two opinions are frequently discussed in terms of one being a government win on liability and the other being a government loss. That framing misses the more important reasoning consistency that unites them. In the Meta case, the change came from rivals and from the evolution of Meta’s own products, as connected content gave way to recommended video and the boundaries between apps dissolved. In the Google case, the change came from an adjacent technology that barely registered as a near-term threat at the liability trial but took center stage by the remedies hearing, and has only accelerated since. One judge let dynamism decide liability. The other let it discipline the remedy. Both judges refused to treat the market as it existed when the government began its work as the market that mattered.
The economics here is straightforward: In sectors characterized by rapid changes and dynamism, whether from significant new entry, converging product features, or large swings in where consumer attention flows, a market share measured at the outset of an investigation cannot simply be assumed to establish market power five years later. Both cases took roughly five years from complaint to decision. Over that span, the products at issue in the Meta case changed their core function, and the products at issue in the Google case faced an emerging class of AI-based competition. Where entry, substitution, and/or converging product features erode an incumbent’s position, an enforcer who freezes the analysis at filing risks overstating the durability of market power and prescribing relief for competitive conditions that no longer exist. In innovation-intensive sectors, that mistake can obstruct the very competitive developments the law should protect.
The burden of proof belongs to the party asserting that another party has durable market power. Historical evidence must be tested against subsequent developments before it can establish present power or justify prospective relief. Remedies must be sized to the competitive landscape the court can actually observe. Judge Mehta and Judge Boasberg applied that standard with full trial records in front of them and reached results tailored to what those records showed.
The Lesson for Ex Ante Regulation Is: Don’t!
That standard is also an indictment of the regulatory alternative. Proposals modeled on the EU’s Digital Markets Act, including the American Innovation and Choice Online Act (AICOA) reintroduced this June as S. 4746, dispense with case-by-case market definition entirely. They designate covered platforms by user, and revenue thresholds set at a fixed moment, impose a predetermined list of prohibited conduct, and provide no clear mechanism to revisit those rules as the competitive frontier shifts. If two experienced judges with hundreds of pages of trial evidence concluded they needed to update their analysis in real time to avoid error, a highly prescriptive statute first drafted around the market of 2021 and revised somewhat in 2026, enforced by rote in 2027, has no chance of keeping pace. Ex ante rules would bake in yesterday’s market structure by design. The courts have spent the past five years showing why that design fails.
The damage would be greatest in artificial intelligence, where the competitive frontier is moving fastest. The bill’s definition of an online platform expressly includes operating systems and digital assistants, along with any service that “enables user searches or queries that access or display a large volume of information,” language broad enough to reach a generative AI assistant or answer engine. Because the user counts of a covered firm’s “similar platforms” are aggregated, an AI assistant that a covered firm launches to serve use cases substantially similar to its search engine could inherit covered status the day it launches, while a rival offering the same product would face no such limits.
Moreover, adding AI to an existing covered platform creates exposure under nearly every prohibition in the bill. For example, an AI-generated answer that draws on the firm’s own maps, shopping, or travel services could be challenged as favoring its own offerings in ranking and presentation. A covered firm is also barred from using nonpublic data generated by users’ interactions with business users’ products to support its own competing offerings, a restriction that bears directly on how assistants are trained and improved. Safety measures receive no special protection. A firm that limits third-party access to a model on safety or security grounds must prove necessity by clear and convincing evidence, supported by records created when the decision was made, and must show that no less anticompetitive means existed. With penalties of 1 to 10 percent of total U.S. revenue at stake, a predictable response is to delay or withhold AI features from American users, as Apple did in 2024 when it held back Apple Intelligence and two other new features from EU customers, citing the DMA’s interoperability requirements.
Caution Is a Good Default In Dynamic Spaces
Both cases remain on appeal. Google noticed its appeal in January 2026 and the government cross-appealed in February, and the D.C. Circuit will weigh in on the timing question the FTC raised in the Meta case. But the district court opinions already stand as two of the most thorough judicial treatments to date of how antitrust analysis can handle markets that refuse to hold still. Their shared answer is caution: skepticism toward market definitions that exclude the products consumers actually substitute, insistence on current evidence rather than a snapshot from the investigation’s opening day, and restraint in imposing remedies when the sector is visibly reorganizing itself.
That is the right default for dynamic markets, and it is the standard against which every proposed intervention in dynamic and frontier sectors, judicial or legislative, should be measured.
