“Google is a Monopolist” – Wrong and Right Ways to Think About Remedies

District judge Amit Mehta’s August 5 opinion in United States, Colorado et al. vs Google that “Google is a monopolist” in search and “has acted as one to maintain its monopoly” in violation of Section 2 of the Sherman Act is a thoughtful, careful, and ultimately straightforward application of the Microsoft precedent. In the opinion’s 286 pages, there is no concession to a “political economy” view of antitrust, and in truth no need for it, as the anticompetitive contractual restrictions it describes are as egregious as it gets. The European Commission’s Android case, resolved in 2018, was based on a similar theory of harm, though bizarrely the Commission never touched the Apple side of OEM agreements, but rather remained focused on dealings with Android OEMs.

Given where US antitrust posture towards Big Tech stood just five years ago, Judge Mehta’s opinion is a spectacular outcome. Google will appeal, though the Judge pushed back hard on all the standard “law & economics” arguments made by the defense (“exclusive contracts are competition for the market,” “rivals could have outbid,” “efficiencies”) which he patiently listened to in all their inane glory. The opinion repeatedly says “market realities matter more than what is theoretically possible” (p.207), and “what Google says has intuitive appeal, but it does not reflect market realities” (p.239). The bottom line, concludes the Judge: “the key question then is this: Do Google’s exclusive distribution contracts reasonably appear capable of significantly contributing to maintaining Google’s monopoly power in the general search services market? The answer is ‘yes.’” (p.216). Amen.

While this is a monumental achievement for the Department of Justice and the States, and the judgment will be pored over for its implications for other Big Tech cases, the focus now is on what measures will be ultimately imposed at stage two of the process: the “remedy trial” whose timetable is being settled before the same judge in early September. The early European cases against Google in search (Shopping 2017, Android 2018) fell miserably at that hurdle: in the excitement of just finding against Google, there was no anticipation that the real mischief would be in remedy design, which was left to Google and led to a dance of egregious uselessness and endless delays, gloriously taking European regulators down the garden path. Indeed, nothing has been achieved since by those remedies – Europe pivoted to ex ante regulation instead.

What remed(ies) would now stand a chance of establishing competition in search, after years which have cemented Google’s durable monopoly? The basic rule for antitrust remedies is they must address the anticompetitive conduct directly and specifically, and the Judge is clearly very focused on the role played by exclusive contracts for default placement. It is also widely expected the DOJ will ask for “structural relief” – generally anticipated to mean a carveout of Android from Search/Chrome. There may well be also a range of disparate proposals put forward by special interest groups (as was the case for Shopping in Europe), all vying for their special corner, and there will be especially strong calls for using the opportunity of the remedy to do something about Google’s data advantage in the realm of AI.

But what remedies will stand a chance of creating actual competition in Search? Because this is ultimately the goal, even though some of us feel that “curbing power” may well be worthwhile in its own right. The reality is that Search sits at the heart of a vast and complex ecosystem connecting multiple parties – users, advertisers, publishers, browsers – and remedies have to approach it as such. There is also the adtech trial in another suit brought by the Justice Department starting in early September, and the parallel Texas case to be tried in 2025 – how these enforcement initiatives will play out in the effort to address Google’s problematic ecosystem remains to be seen.

But there are (at least) two ways of getting remedies wrong: picking one “big” obvious remedy and hoping it will solve all problems; or putting forward a “Rube Goldberg machine” of complex ideas trying to address the issues of the whole ecosystem, which will not get off the ground. While all ideas should be vetted at this stage, we discuss below what might and might not work, bringing together our experiences as a technologist and an antitrust economist. None of this pretends to be definitive, and it is certainly not easy.

Wrong 1: Pick one “big obvious remedy” and hope this will introduce enough competition to solve all problems

The obvious candidate for remedy is prohibiting Google from entering into any agreement paying a device manufacturer for exclusive default placement of Google Search. The Judge is very clear that these agreements have killed incentives for others to invest in a search engine, and kept Apple in check as a potential competitor. And as framed – anticompetitive conduct perpetrated through highly objectionable contract terms imposed by a monopolist making payment for exclusive default placement – the case indeed appears to have a clear and easy answer: remove the ability to pay for exclusivity. As the judge anticipates in the opinion, “[w]hat’s more, a non-exclusive default would still provide all the convenience and efficiency benefits that Google touts” (p.249).

How far will this go in encouraging adoption of other engines though, and competition? Apple and other OEMs would lose the exclusivity payment, but how does this change the picture for adoption and innovation? Google will most likely be prohibited from making payments for exclusivity; but what if Apple (and others) were to take the view that devices need a pre-installed default anyway (users expect it out-of-the-box) and as no other engine is remotely of comparable quality, they will still pre-load Google Search on their devices? The Judge can direct remedies at Google, but it cannot tell Apple what to do as part of this case. He could go further than outlawing payment for exclusivity, and outlaw any payment for pre-installation. But he cannot tell Apple not to pre-install, and one can anticipate whether some form of payment could still be worked out – for instance some form of commission on revenues from search queries coming from Safari, essentially some variant of the existing search royalties system.

Note that this would be a rational business decision for OEMs that are not party to the case, which exposes the reality that there can be no fundamental change to market structure in Search arising from amending a contract. It might have made a difference 12-14 years ago, but not today. There are many ways to shape user behavior, and while the big players often practice circular firing squads, their interests are also often aligned and they will figure out how to make that work even if it's less effective than defaults. So while there is an obvious remedy that addresses the conduct, it seems hard to see how it can create competition. We need more. How much more, and how?

Wrong 2: Build a Rube Goldberg machine of multiple remedies that tries to address all Google Search’s monopolistic behavior at once

Google Search has woven its tentacles deeply into the structure of the web, to the point that the web is in many ways Google’s other walled garden. Publishers get a huge portion of their traffic from Google, often the majority. Google Search advertising revenue is a massive share of advertising in general. Payment from Google finances over 90% of the web browser market. And that’s without mentioning commerce or Android.

Trying to architect the entire ecosystem with a set of detailed problem-by-problem remedies that operate together in an integrated manner is too complex to even be attempted. Google had both years and billions to spend turning the web into a network of offers you can’t refuse. No remedy system of matching complexity can reasonably be devised, and if someone somehow managed to create it, it would almost certainly break within months.

A better way: remedy individual points of control

Dave Clark (of “rough consensus and running code” fame) devised an approach he called control point analysis to investigate how a given technical system will give some actors greater power by virtue of “control” over key components of that system. For instance, before someone can retrieve a web page they’ll need to have selected a device, an operating system, a network provider, and a browser – steps that already involve control points since the device likely comes with an OS that defaults to a given browser, etc.. Then a series of further steps needs to be carried out: finding a URL for that page, extracting and resolving the domain name, connecting to the server, fetching the content of the page over HTTP, interpreting the content of the page, rendering it and its embedded elements. This introduces additional opportunities for control: the URL might be found via search, the domain name needs to be registered and resolvable, the browser could prevent some resources from loading (as with ad blockers), or it could refuse to cooperate with a server’s requests (as happens with cookie protection). Note that some control points can be primarily valuable for observation purposes: for instance, a domain name resolver that would commonly refuse to resolve domains would end up replaced with another so its opportunity for influence is limited, but knowing what domains people are looking up (from which IP address) is valuable data.

Control points are largely created by a system’s design, and Judge Mehta won’t be able to change the web’s architecture. However, remedies can be targeted at specific control points in order to diffuse rather than concentrate power, and that is the approach we advocate for: list the many ways in which Google Search connects to the rest of the ecosystem, identify how those specific connections give them power over the web, and devise independent remedies for each. Individual remedies can be relatively orthodox, but the guiding approach to how they are arranged together can be novel. Again, none of this is simple.

In a sense this is a typical technologist solution: modularize the problem and solve the modules independently. It may or may not be useful to break Google up, but we can break up the problem no matter what. As we have no “silver bullet” from the perspective of antitrust either, a selection of possible interventions is where we need to start. Of course going through every way that Google exerts power over the web would make for a much longer piece than would fit here, so below we list only a few candidates. Others may have a different list, and each item will require serious further thinking, including from the perspective of legal feasibility.

Example 1. Can choice screens be part of the solution?

What about leaving it to users and pairing a prohibition on exclusivity (or even preinstallation) with a “choice screen” obligation, in which users make their own choice? Choice screens through which users pick a search engine seem like a must for a competitive search market.

The first question, however, is: how does one bring that about? An obligation on whom? The judge can force this on Google and therefore on Android devices as well as Chrome, but how can it force an OEM like Apple to adopt a choice screen in this case? The European Commission mandated it in the Android case for Android devices, but famously did not go near Apple devices. Regulators in Europe are moving forward with choice screens under the Digital Markets Act (DMA), while the UK is exploring the same under the Competition & Markets Authority’s Market Reference.

But again how would Judge Mehta have jurisdiction telling Apple what to do here? An independent OEM that believed choice screens do not work – because they believe, rightly or wrongly, that other engines are of lower quality and users would pick Google anyway – naturally has no incentives to go that way. Perhaps the judge could prevent Google from entering into any search royalties deals that do not involve a choice screen, so that the OEM or browser vendor could only be paid by Google if it went this route. But it is unclear whether this is possible, and whether it could be worked around by the parties in some way.

The second question is how effective would a choice screen be at fostering competition? Choice screens can be designed to be amazingly ineffective, as Europe has repeatedly shown the world. The experience of Android has been an outright failure, and early evidence from implementation of the DMA on Apple devices shows there is some movement to other engines, but minor. That said, recent work from Mozilla makes the case that they can also be designed effectively. There is also academic work on this – see inter alia “Randomization as an Antitrust Remedy” by Francesco Ducci, and a skeptical analysis by Megan Gray in Tech Policy Press, among others. It’s also important that such choice screens be called upon not just once but everywhere that default search can be set. The design of such a screen should be carried out and maintained over time through a multistakeholder standards process that reflects the needs and concerns of the wider internet community.

One persistent challenge with choice screens is that search engines are hard for users to compare and have properties that are not readily evident even when experienced, such as how they manage user data. Evidence from Mozilla’s study is that users actually prefer it when the choice screen provides them with more information. (The study was on browsers, but the findings translate well.) The group in charge of maintaining the choice screen should therefore be able to surface a number of criteria – for instance privacy or journalism-friendliness – that people can use to filter their decision on. This could help dampen the benefits accrued to Google’s brand recognition by years of monopolistic practices.

In order to be eligible to feature on the choice screen, a search engine must agree to specific terms that include participation in the royalties system that funds browsers (about which more below). For instance Google could be allowed to participate in that choice screen and only allowed to pay royalties to browsers and operating systems that rely on it. This might draw other participants into the agreement even if they aren’t otherwise compelled. Will this work to create competition? We cannot tell, but worth experimenting with.

Example 2: Interoperability and a touch of collective bargaining for publishers

One of the ways in which Google has harmed the ecosystem is by imposing terms on publishers about how they are to make content available (AMP, schema.org, Core Web Vitals, etc.), extracting ever more coerced labor while continually decreasing the volume of traffic provided in exchange for content to index by preferencing links to its own properties or responses that keep users on Google’s search engine results page (SERP).

They are able to exert that power because they have monopolized a critical control point: how content is discovered. Should publishers wish to resist a change taking more from them in exchange for less, Google only has to mutter “nice traffic you got there, shame if anything happened to it” for everyone to fall in line. And needless to say, this advantage can readily compound into AI to prevent publishers from being tempted to exclude their content from LLMs.

A remedy here might involve imposing interoperability terms (e.g., content can only be required to be HTML as defined in such and such specification) along with purpose limitations (for instance, that content obtained this way may be used for search results with links but not LLMs). But that should both lock down innovation potential, and still be gameable. We can find a better approach with a less restrictive understanding of interoperability. Interoperability is often considered in terms of a specific format, protocol, or API that all actors must use, but in the real world interoperability, at least for technology of any real complexity, isn’t a state – it’s a process. An interoperable system will in practice be tied to a group of people with a specific power dynamic who maintain the interop of the system and deliberately shape what it can and can't do.

Taking this into account, the remedy could designate the group in charge of this interoperability mandate and purpose limitations. Groups that manage interoperability and paraphernalia are standards-setting groups, and there is an abundance of existing regulation and precedent to govern how such groups operate, including unrestricted participation, a consensus process, and objective criteria for selecting options. With the right charter and oversight, such a group could effectively create a form of permanent collective bargaining for publishers and place limits on self-preferencing without requiring new legislation. It would get ugly and political, but these things are supposed to be ugly and political.

Example 3. Data advantages

A central component of the conversation around Google’s monopolistic behavior in search revolves around its data advantage. Competitors such as Microsoft sought support from European regulators in addressing the data advantage from the early days of the Shopping case – to no real avail. But calls to deal with the “advantage” are especially acute in the age of AI, given the role of data in model training.

The advantage can be broken down into two components. First is the advantage that Google gets via the data it obtained by operating its search product as the dominant player. This part requires a nuanced approach. In general, there is no harm from a service or publisher learning from its operations. This leads the operator to understand its audience better, which would typically lead to higher quality of service and is part of a business’s typical competitive toolbox. A key question here is: is this an advantage that mechanically tips the market such that the moment one search provider grows slightly larger than the others it will lead to a runaway growth cycle with no diminishing returns, ending in insuperable dominance? That seems unlikely, even accepting that long-tail queries may be highly valuable. If it were, then after it got larger than its competitors there would be no reason for Google to spend so much paying off browsers and hardware makers to be the default with the eagerness of an anxious mobster buying judges and witnesses.

While that advantage may not be intrinsically anticompetitive, it is proportional to market share. As established in court, Google’s market share was achieved illegally, and we should not let a monopolist retain ill-gotten gains, especially if they can be used to maintain a strong market position. A remedy here would be to consider full disgorgement of the data that Google obtained from operating search, including all downstream assets like trained models, followed by a period of several years during which the allowed retention period the new data Google obtains from operating search would follow an equation of the form: 18 months x (1 - 0.market-share). So at 95% market share Google would be allowed to retain new data for 18 x (1 - 0.95) = 0.9 months, or about 27 days, while at 30% market share they would be able to retain new data 18 x (1 - 0.3) = 12.6 months, which is a year and change. (The exact details may of course vary.) Such a proportional approach would ensure that Google is penalized only so long as its market share remains at a level caused by its monopolistic behavior.

The second aspect is the advantage that Google gets from the data it extracts from countless control points outside of search and that it may conjoin with search data in order to gain an edge over its competitors. To give an example, whereas other browsers like Apple’s Safari will synchronize a user’s browsing history across devices using end-to-end encryption that guarantees that the browser vendor cannot observe that history, Google’s Chrome employs an array of deceptive design patterns to trick people into agreeing to hand their entire browser history over to Google in ways that can’t even be blocked with an ad blocker. Having the full browsing history of three billion people can provide a significant advantage to a search engine – and Chrome is only one of its products.

This part of Google’s data advantage requires considerably less nuance. It encourages a cycle in which search profits are used to subsidize products to below market prices – which others can’t fairly compete with – in other markets that have observable control points only to obtain data that bolsters search (and advertising). This distorts both search and other markets, and creates significant and wholly unnecessary privacy risk for people. The remedy on this chunk can be clear-cut: full disgorgement and never do it again.

Example 4. Browsers as critical infrastructure

Browsers (and browser engines, the software component that renders web content inside a browser window) are an incredibly valuable component of the wider digital ecosystem. In many ways they offer key protection against an even stronger dominance over our digital lives from app stores because they are a separate, relatively open, and highly capable distribution vector. Browsers are also tremendously complex things. A widely accepted number in web circles is that maintaining browsers at their current level costs about $2bn per year. Even if the exact number may be an exaggeration, it's at the right order of magnitude. This puts funding browsers far outside the reach of open source infrastructure support systems like Germany’s Sovereign Tech Fund. And security failures in a browser have very serious immediate impacts on the five billion people who use them daily. We need to maintain browsers, and it's expensive.

Where does that money come from? Almost entirely from search. Mozilla Firefox and Samsung Internet have all been paid for by Google Search, as has Google Chrome, though that is internal to the company. Apple Safari is part of the Apple integrated ecosystem but Google’s exclusivity payments have contributed billions to that. This group already covers all three major browser engines. Microsoft Edge, DuckDuckGo, and Brave all have their own search engine tied to the browser. Apart from a handful of exceptions among the smaller browsers, search is how browsers are paid for.

This funding mechanism has been problematic. Not only is it key to artificially maintaining Google’s dominance in search, it has also allowed Google to directly influence design decisions made by rival browsers, for instance delaying the implementation of privacy protections or preventing browsers from correcting some of Google’s more egregious user experience decisions (ads that look like results, AMP pages that pretend to be another, more trusted website, etc.).

A remedy for the search/browser relationship needs to maintain funding for browsers while addressing the control that it grants Google over our shared browser infrastructure. But again Judge Mehta can only impose terms on Google and not on other market participants. However, Google is such a significant source of revenue for others that by constraining the terms of the deals that Google is allowed to enter into, he could constrain the behavior of those who wish to deal with Google.

We already discussed banning payments to be the search default. Whoever has the default has the market, and whoever has the market can pay more than others. A search default is not conducive to competition. Then, because browsers need funding, one could allow search royalties in which the browser gets a cut of every search query’s revenue. But this requires controls: Google could still pay higher royalties and browsers could still decide to default to Google for other strategic reasons.

Finally: Breaking the company up

Given the complexities of any remedy (and the list above is a clear illustration), the question needs to be whether “structural relief” – i.e. separation of parts of the ecosystem, notably Search (and perhaps YouTube and Google Maps) from Android/Chrome, with the potential for separate ownership – could be preferable as a ‘once and for all’ solution.

The point of breaking up a company is of course to address incentives which fundamentally drive conduct towards exclusion and self preferencing – or, in the language we have used in this piece, to ensure that different “points of control” cannot be used by the same entity to exert greater power. (A good discussion is contained in “Breaking Up Big Tech: Scissor Line Suggestions for Smart Cuts,” a May 2024 paper by Maarten Pieter Schinkel and Ruben van Oosten.) One view of breakups is that “it is all too difficult and will require years of complex litigation, so what’s the point?” Another is that “it could be counterproductive if it meant loss of economies of scale and scope” (but frankly we are way past that point). Yet another is “that this is the only approach that could realistically work to tame a sprawling ecosystem, and in fact break ups are frequent in business and even beneficial to shareholders as they create value.”

All of these are simultaneously true, but that does not mean we should not explore them. Break ups are challenging, but they seem also highly justified in this case. We should not care about the loss of economies of scale for Google if the remedy unleashes innovation and competition from third parties. We will see how the DOJ will frame its anticipated plea for structural relief in the remedy trial. But after all, setting the default for Chrome is not very different from setting the default in Safari, the difference is that we have an agreement between two completely different products inside a company instead of with another business. And the deals that Google makes with OEMs to obtain the default in Android are not essentially different. The prospect of divestment of Search from Chrome and Android makes getting the royalties system right all the more important.

Independent Chrome and Android entities would be a major win for an open internet. Google uses those control points to push its own solutions in multiple areas such as identity, email, payments, mapping, and more, as explained clearly in Konrad Kollnig’s article on the central role that Chrome plays in Google’s monopolistic behavior. Independence – if they can keep it, hence the importance of the royalties scheme – would have the knock-on effect of opening many other markets up for competition and innovation.

Another interesting break-up – which could be complementary – would be to split the index, which is to say the part of Google Search that scrapes the web and makes that content searchable, from the search user interface, and manage that index as a public utility that different search services could rely on and pay for, an idea that was suggested in a recent paper.

While perhaps harder to justify within the bounds of case, such a split would present a major benefit for democracy. Search engines are very similar to news media: they have an editorial function that determines what is relevant on any given topic. The interaction modalities differ but the effect on the information environment is comparable. With effectively only one editorial line determining relevance in search, the internet has a Pravda problem. Maintaining a shared index would make it a lot easier for alternative ranking and filtering — other editorial lines, that is — to be produced, in the same way that the emergent darling social media service Bluesky has open algorithmic feeds and composable content moderation. Browsers could access multiple search services sitting atop the same index but with different ranking choices simultaneously. This would effectively bring us multihoming in search. The technological lift would be significant, but not impossible.

Overall, this week’s ruling provides a unique opportunity to try and “reset” at least in part a Search environment which had become impossibly skewed, supporting the growth of Google into an unassailable ecosystem. If antitrust enforcement is to be more than a specialty sport, with its own rituals and arcane mores, it needs to show that it can do more than issue decisions, opinions, and judgments finding ecosystem owners had been somewhat naughty, and then let it all drift forward into the distance. It needs to use the leeway of these foundational moments to scope out ambitious interventions that stand a chance of allowing challengers to come forward. That Big Tech actors themselves do not seem to be worried – negotiating similar deals around AI right as this case was being argued – is a sign that they anticipate a weak outcome. Targeting each control point as we advocate doing may produce a relatively large list of remedies to deploy, but that is proportional to the number of tentacles that Google Search has slithered into our digital ecosystem. The difficulties and resistance will be huge – but the failure of prior efforts in Europe should suggest not that intervention is futile, but that only an aggressive posture (like some of the remedies we suggest here) stands a chance of making any difference.