A judge dismissed Chegg’s and Penske Media Corporation’s antitrust lawsuits over Google’s AI Overviews on Wednesday. The decision did not establish whether AI-generated search answers cost publishers readers. We believe Congress should give publishers an enforceable choice about whether their content appears in those answers without forcing them out of ordinary search results.
That position does not assume the companies proved their alleged losses, or that Google owes publishers a payment for every search result. It recognizes a problem the lawsuits could not settle: who gets to decide how a publisher’s work is used in AI search.
What the judge decided
Chegg and Penske Media Corporation alleged that Google used publisher content in generative-AI products without payment and that AI Overviews diverted visits and revenue from their sites. Google denied wrongdoing. In court, it argued that it had never agreed to provide referrals in exchange for publishers’ content and was not required to index that content on their preferred terms.
U.S. District Judge Amit P. Mehta dismissed the companies’ federal antitrust claims because their amended complaints did not plausibly allege an agreement or make other showings those claims required. As Engadget reported, Mehta wrote: “But an expectation is not an agreement. It is simply how a general search engine works”.
The distinction matters. Publishers may have built businesses around visits from search, but an expectation of those visits does not, by itself, create the agreement alleged in these cases. Mehta also declined to hear the companies’ state-law unjust-enrichment claims and dismissed them without deciding their merits. His order is appealable; the available reporting does not establish that either company has appealed.
A pleading-stage dismissal is not a measurement of traffic. It tells us these complaints did not satisfy the requirements for the antitrust claims they brought. It does not tell us whether AI Overviews has reduced Chegg’s or Penske’s revenue.
What the traffic evidence can and cannot show
There is evidence worth investigating. In a preregistered field experiment involving 1,100 participants, researchers reported that removing AI Overviews and AI Mode increased click-through to publishers. Their paper was posted as an arXiv preprint, not presented here as peer-reviewed research. The result concerns that experiment, not every query or the losses alleged in these lawsuits.
Google, for its part, says its generative search features benefit users and websites, including through greater user satisfaction and search activity. Those assertions are not independently verified by the material available here. They also address different questions from whether particular publishers receive fewer valuable visits.
The strongest objection to a publisher-control rule deserves more than a shrug. People may value answers that spare them a trip to another page, and a search service cannot guarantee a publisher a stream of visitors. Google should be able to improve its product without treating every change in referral patterns as a breach of an unwritten bargain. Mehta’s ruling makes that legal boundary clear.
But that argument does not settle whether publishers should have a say in a distinct use of their content. The choice is not between compelling Google to preserve old traffic levels and allowing publishers no meaningful control at all.
Free newsletter
Get the morning briefing
Start each day with the AI stories that matter and why — a short, free email from our newsroom.
Why a test is not yet a safeguard
In June, Google said it was beginning to test a Search Console control letting site owners exclude their sites from generative-AI Search features. Google said opting out would end traffic and impressions from those features but would not affect rankings elsewhere in Search. That is the company’s description of a test, not an independently verified account of how the control works or evidence that it is broadly available.
An opt-out necessarily involves a trade-off: a publisher that declines inclusion cannot expect visits from the feature it leaves. The critical question is whether it can make that choice without also losing its place in ordinary search. Publishers should not have to accept one use of their work to remain eligible for another.
The United Kingdom has taken a different route from these U.S. lawsuits. On June 3, its Competition and Markets Authority imposed a publisher conduct requirement on Google covering controls over use of content in generative-AI search features, transparency and attribution. That requirement is not a U.S. court remedy, nor does its adoption prove it is the right model in every detail. It does show that publisher choice can be addressed as a rule for search conduct rather than squeezed into a claim about an alleged agreement.
The choice Congress should protect
We think Congress should establish an enforceable right for publishers to exclude their content from generative-AI search features while remaining eligible for ordinary search indexing. The rule should require clear attribution and publisher-facing information about when content appears in those features and what referrals follow. That is one framework for an informed choice, not a mandate that every AI answer generate a payment or a click.
Lawmakers should leave room to assess how such a rule affects search users and publishers alike. They should also require evidence about referrals rather than assume either Google’s account of user benefits or the plaintiffs’ account of lost revenue is complete. The 1,100-person experiment gives that inquiry a starting point, not a final answer.
Mehta was right to judge the complaints against the law they invoked. Congress has a different responsibility: to decide whether publishers can refuse a new use of their work without disappearing from ordinary search. It should protect that choice directly.