Understandably, the headline suggestion has been to break up Google, perhaps by forcing it to split off its Chrome browser or Android mobile operating system—both of which have been key to shoring up Google’s search monopoly. But the DOJ also proposed changes (or “remedies” in antitrust-speak) that could severely crimp Google’s efforts to maintain a leading position in the age of AI-infused search.

A rapidly evolving AI search landscape

The department’s suggested fixes would deprive Google of some of the data that it can easily draw on today when delivering AI-powered search results, and perhaps shape the economics of how AI search works.

The field of AI search is still in its infancy, but is rapidly gaining pace.

Google has been incorporating “AI overviews” into its search results since earlier this year, combining the chatbot functionality of its Gemini AI model with a technique known as retrieval-augmented generation (RAG), which retrieves fresh information from websites.

The idea here is to provide more up-to-date responses than could be generated purely through a large language model that was trained at some point in the past, and also to avoid the “hallucinations” that are endemic to LLMs—though Google’s AI overviews had an embarrassing launch, with people quickly finding that the overviews sometimes drew on parodic sources to earnestly recommend that people eat rocks or put glue on their pizza.

Meanwhile, Google also faces new entrants in the search market, like You.com, Andi, and Perplexity AI. Perplexity will later this month start running ads next to its AI-generated search results. (Full disclosure: Fortune has a partnership with Perplexity.)

“Artificial intelligence—while not a substitute for general search—will likely become an important feature of the evolving search industry,” the DOJ wrote in its proposals to the court. “It is, therefore, critical that any remedy carefully consider both past, present, and emerging market realities to ensure that robust competition, not Google’s past monopolization, will govern the evolution of general search and text advertising.”

DOJ recommendations: Opt-outs from AI training and RAG and prohibitions on exclusive partnership terms

The department made two specific suggestions that it claimed could stop Google from using its existing search monopoly to ensure its ongoing dominance as AI search becomes ubiquitous.

The first involves Google’s web crawlers, which the company has used since the last century to index web pages and their contents so it can include this information in its search results.

Website publishers clearly benefit from that inclusion, but they can use a file called robots.txt to tell Google’s web crawlers to stay away. Just over a year ago, Google made it possible to use this file to also allow or block the training of Google’s AI apps on a site’s content. Following publisher feedback, Google said in February that opting out of feeding Google’s AI efforts wouldn’t affect a site’s appearance on Google Search.

However, the advent of Google’s AI overviews complicates this picture—it’s a Google AI service, but also part of Google Search. The DOJ seems to think that, with Google’s search monopoly leaving content publishers with “little to no bargaining power,” the Big Tech firm should be forced to “allow websites crawled for Google Search to opt out of training or appearing in any Google-owned artificial-intelligence product or feature on Google Search such as retrieval-augmented-generation-sourced summaries.”

Fortune has asked Google to clarify whether it is possible for a publisher to reject the use of its content in RAG summaries while still maintaining a presence in Google Search.

The DOJ also proposed prohibiting Google from “using contracts or other practices to undermine rivals’ access to web content.”

Earlier this year, Google struck a deal with Reddit that was worth a reported $60 million. Under the arrangement, Google gets to use posts from the online discussion site to train its AI models and feed into Google Search. A few months later, it became apparent that only Google could directly surface content from Reddit, with rival search engines no longer being able to provide useful links to that content.

This may be a financial boon to Reddit—the deal’s announcement shortly before Reddit’s IPO certainly made for fortuitous timing—but it clearly also disadvantages Google’s rivals. In the DOJ’s words: “Google’s ability to leverage its monopoly power to feed artificial intelligence features is an emerging barrier to competition and risks further entrenching Google’s dominance.”

Google warns of ‘unintended consequences’

For its part, Google responded to the DOJ’s proposed remedies by warning of “unintended consequences.” On the AI points specifically, regulatory affairs VP Lee-Anne Mulholland said in a blog post that “hampering Google’s AI tools risks holding back American innovation at a critical moment.”

“Business models in AI, much less winners and losers, have yet to be determined, and competition globally is fierce. There are enormous risks to the government putting its thumb on the scale of this vital industry—skewing investment, distorting incentives, hobbling emerging business models—all at precisely the moment that we need to encourage investment, new business models, and American technological leadership,” Mulholland said.

The DOJ will submit more detailed proposals to the court late next month, with Google getting the chance to submit its own preferred remedies by Dec. 20. The two sides will argue their positions in an April trial, with a final ruling following later in 2025.

Google is also appealing the underlying antitrust ruling, so it remains to be seen how quickly the final remedies become a reality, or if any of them will ever be implemented.