Most search strategy decks written in the last eighteen months contain an unstated assumption, usually sitting in the risk slide: that somebody above the marketing team's pay grade will eventually fix this. A regulator, a court, a settlement. On October 1 a federal judge removed one of the larger candidates from that list, and did it in language that is worth reading rather than summarizing.
The AI Overviews lawsuit brought by Penske Media, the publisher of Rolling Stone, Variety, Billboard and The Hollywood Reporter, was dismissed in full. So was the near-identical action brought by Chegg. Judge Amit P. Mehta of the US District Court for the District of Columbia heard the two together and disposed of both in one memorandum opinion running 41 pages.
The headline most outlets ran was that Google won. True, and not the interesting part. The interesting part is the reasoning, because it describes the actual legal status of the arrangement that every content program in the world has been built on top of, and the answer is that there was never an arrangement.
Five for five, with nothing left standing. A plaintiff with Penske's resources and a well-documented injury did not lose on one technicality, it lost on every theory it brought, which tells you the problem was structural rather than tactical.
What the AI Overviews lawsuit actually claimed
The AI Overviews lawsuit argued that Google used monopoly power in general search to coerce publishers into supplying content for free, then used that content to generate AI answers that keep the reader on Google instead of sending them onward to the site that produced the material.
Penske filed on September 12, 2025, and framed the position as a forced choice: allow Google to crawl and be cannibalized, or block Google and disappear. Chegg brought substantially the same case from the education side. Both argued that the exchange underpinning the web for twenty-five years, free crawling access given in return for referral traffic, had been unilaterally rewritten by the party with all the leverage.
The complaint leaned on third-party research rather than invented numbers, which is the right instinct and worth copying. It cited projections of advertising revenue loss and traffic decline in the 20 to 60 percent range, and research finding that AI Overviews cut click-through by as much as 34.5 percent for the top organic result.
| CLAIM BROUGHT | WHAT IT REQUIRED THE PLAINTIFFS TO SHOW | WHY THE COURT REJECTED IT |
|---|---|---|
| Reciprocal dealing | An actual agreement that Google would send referral traffic in exchange for crawl access | The plaintiffs pleaded an expectation of traffic, not negotiated terms, commitments, or communications showing mutual assent |
| Tying | Two separable products, with access to one conditioned on acceptance of the other | The court did not find the conditioning relationship the theory needs between general search and the AI features |
| Unlawful monopoly maintenance | Antitrust standing in the general search services market and anticompetitive conduct sustaining the monopoly | The plaintiffs were found to lack sufficient antitrust standing in that market as publishers rather than search competitors |
| Attempted monopolization and monopoly leveraging | A clearly defined publishing market plus a dangerous probability Google monopolizes it | The publishing markets were not clearly defined, and the likelihood of monopolization was not adequately shown |
| Unjust enrichment | A benefit conferred on Google that it would be inequitable to retain | Fell with the rest once the underlying bargain was found not to exist as an enforceable obligation |
Read down the third column and a pattern appears. Four of the five failures are about definition and standing rather than about whether Google did the thing. The court was not asked to decide, and did not decide, that AI Overviews are harmless.
Why the AI Overviews lawsuit failed on a single word
The AI Overviews lawsuit collapsed on the difference between an expectation and an agreement, a distinction the court drew in one sentence that is now the most quotable line in search law: publishers pleaded only that they expected traffic in return for free content, and an expectation is not an agreement.
That sentence does a lot of work. Every SEO engagement ever sold rests on the premise that good content, made crawlable, earns traffic. The industry treats that as a deal with terms. The court looked for the terms and found none: no negotiated commitments, no communications showing mutual assent, no meeting of the minds. What exists instead is a convention, honored for decades because it suited both parties, with nothing underneath it.
The court put the same point more bluntly elsewhere in the opinion, describing automated web crawling and publisher expectations of search referral traffic as reflections of general search engine functionality rather than an enforceable or coercive bargain. Functionality, not bargain. A robots.txt file is a request that a well-behaved crawler chooses to honor, and it was never a contract.
“An expectation is not an agreement. Twenty-five years of content strategy were built on the first one while assuming the second, and a federal court has now written down which of the two actually existed.”
The practical read is that litigation over referral traffic has to clear a bar nobody has yet cleared, and the bar is not about proving harm. It is about locating an obligation. Until somebody finds one, or a legislature writes one, the crawl-for-clicks exchange stays a courtesy.
The court agreed the harm is real and dismissed anyway
Judge Mehta wrote that the court does not treat the plaintiffs' alleged harms lightly, and is not unsympathetic to the situation publishers now find themselves in, nor to the knock-on consequences for journalists, educators and other online creators whose content Google takes and repurposes without compensation.
That is an extraordinary paragraph to find in an opinion that then dismisses every claim. Courts are not obliged to editorialize. Including it signals that the judge saw the injury clearly and concluded the law as pleaded did not reach it, which is a message aimed at Congress more than at the parties.
Decline figures cited in the publisher litigation and the surrounding reporting, shown as percentages. These are cited projections and third-party measurements, not Something Inc. research
Note what the chart is and is not. These are projections and network-level measurements from different methodologies, lined up because they were the evidence in play, not because they are directly comparable. The honest summary is that credible estimates of referral decline now cluster somewhere between a fifth and a half, and the direction is not in dispute by anyone including the defendant.
Which makes the outcome the clarifying event. If a well-resourced plaintiff with that evidentiary record loses on all five theories, the legal route is not a line item in your 2027 plan. The compensation experiments Google is running on its own initiative, including the small payouts visible in the AI contribution pilot and what it actually pays for, remain voluntary, and nothing in this opinion makes them less so.
What the ruling changes for enterprise search strategy
For most enterprise search programs the ruling changes no tactic at all, and changes one planning assumption completely: the probability that referral volumes get restored by somebody else should now be set at approximately zero for the planning horizon you can actually budget for.
That sounds bleak and mostly is not. Teams that already shifted from measuring sessions to measuring presence in answers have been operating as if this were true for a year. The teams this hurts are the ones running a traffic-recovery narrative internally, telling a board that the dip is temporary and regulatory pressure will correct it. That story is now harder to tell with a straight face.
| PLANNING ASSUMPTION | STATUS AFTER THE DISMISSAL | WHAT TO DO ABOUT IT |
|---|---|---|
| Litigation or regulation will restore publisher referral traffic | Materially weaker. Five theories failed at the pleadings stage before any discovery | Remove it from forecasts entirely. Model the current referral baseline as the new normal and plan growth from there |
| Crawl access is an implicit contract we can enforce | Rejected in writing. The court called it functionality, not a bargain | Treat crawl permissions as a business decision you make deliberately, per crawler, rather than a right you are owed something for |
| Blocking AI crawlers is a nuclear option nobody takes | Unchanged, but now the main lever the court itself points to | Audit which crawlers you allow and what each one returns. Decide per surface rather than site-wide |
| Being the cited source inside an answer is a nice-to-have | Now the primary defensible position in the channel | Resource it like the acquisition channel it has become, with its own targets and its own reporting |
| Our zero-click exposure is roughly the industry average | Unknowable from industry averages. Exposure varies enormously by query mix | Measure your own answer-surface share on your own query set instead of importing a benchmark |
The last row is the one teams skip. Published decline figures are network aggregates across wildly different content types, and a B2B software vendor whose queries are comparison and evaluation intent has a completely different exposure profile from a celebrity news publisher. Importing somebody else's 40 percent into your forecast is how a plan gets built on a number that was never about you.
The three levers that survive this ruling
Three levers remain fully under an enterprise's control after the dismissal: what you let crawl you, whether you are the source an answer cites, and whether your measurement describes the channel as it exists now rather than as it worked in 2022.
Start with crawl governance, because the court effectively named it the remedy. The decision of which agents may read your content, and what they do with it, is now a deliberate strategic choice rather than a default. The mechanics got easier this year, and the tradeoffs are laid out in our breakdown of how the content signals policy turned AI input into an opt-out. The honest caveat is that blocking reduces citation eligibility along with training exposure, so this is a dial with a real cost on both ends, not a free win.
Second, citation earning. If the referral click is structurally scarcer, the unit of visibility becomes the mention inside the answer, and the work to earn it is specific and learnable rather than mystical. Our reference treatment of how answer engine optimization actually works covers the structural rules, and the demand-side version of the same problem, where getting named is the whole outcome, is what generative engine optimization engagements are built to produce.
Third, measurement honesty. A program reporting on sessions alone will show decline regardless of how well it performs, which eventually gets good work defunded. Reporting that separates answer presence, branded demand and referral clicks tells you which part moved and why, and it is the only way to defend a content budget in a channel where the click is no longer the deliverable. That separation matters most where content volume is the strategy, a tension covered in our look at content saturation and what it did to content marketing returns.
There is a fourth thing worth naming that is not a lever so much as a posture. Publishers spent two years hoping the arrangement would be restored. The ruling says it will not be restored because it was never codified, and planning from that premise is simply more accurate than planning from hope. Accurate beats comfortable in a forecast that somebody is going to be held to.
Questions teams are asking about the AI Overviews lawsuit
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Tyler leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.