- On September 30, 2026, a federal court dismissed the Chegg and Penske Media antitrust suits against Google over AI Overviews. [1]
- The central claim failed because the publishers pleaded an expectation of search traffic, not an agreement. [1]
- The tying claim failed because the court treats AI Overviews and search as one integrated product. [1]
- The unjust enrichment claims were not decided. The court declined jurisdiction over them. [1]
- The Penske dismissal is without prejudice, in a final, appealable order. [2]
- The same day, Cloudflare put Pay Per Use into beta: AI companies report each use of a page and pay for it. Usage is self-reported. [3]
- Cloudflare’s Monetization Gateway, in closed beta, charges agents per request over HTTP 402. [4]
- Google’s documentation says robots.txt rules for Googlebot are the control for AI features in Search. [5]
Two documents were published on September 30, 2026, and they describe the same problem from opposite ends.
The first is a 41-page opinion from Judge Amit Mehta in Washington, dismissing the antitrust suits that Chegg and Penske Media brought against Google over AI Overviews. [1] The second is a blog post from Cloudflare announcing that AI companies can now pay publishers each time they use a page. [3] One says the law does not recognize the bargain publishers thought they had with search. The other tries to build a bargain out of HTTP status codes.
Neither document gives a publisher what it wants. Together they are the clearest picture yet of where the next few years go.
What the publishers argued
Chegg filed first, on February 24, 2025. Penske Media followed, with a group of plaintiffs that includes Billboard, Deadline, The Hollywood Reporter, Rolling Stone, and Variety. [1] The court handled both sets of motions in one opinion because the complaints, in its words, substantially overlap. [1]
The core theory was reciprocal dealing. The publishers said Google sells them something, search referral traffic, and makes that sale conditional on receiving three kinds of content at no cost: content it can republish in snippets, content it can use to train its models, and content it can retrieve and repackage in generated answers. [1] They pointed out that other companies building AI answers, including OpenAI and Perplexity, pay publishers for content, and that Google does not. [1]
Penske added a tying claim: that Google uses its monopoly in general search to force a second product, AI Overviews, onto users who cannot switch it off. [1] Both sets of plaintiffs also brought monopolization claims and a state-law claim for unjust enrichment. [1]
Why it failed
Reciprocal dealing needs a deal. The court looked for one and found only a habit.
The opinion lists what the complaints did not contain. The plaintiffs, it says, “do not assert that Defendants offered terms they accepted; the parties negotiated quantity, duration, or exclusivity; or either side made any commitment.” [1] What they pleaded was an expectation that Google would send traffic if they made their content available for free. Then the two sentences that will be quoted for years:
But an expectation is not an agreement. It is simply how a general search engine works.
Memorandum Opinion, Penske Media Corp. v. Google LLC, D.D.C., September 30, 2026
This was not a first impression. The same judge had dismissed a similar suit by news publishers in March, and the opinion says the plaintiffs offered no persuasive reason to change course. [1]
The tying claim fell for a reason every SEO should read twice. A tie requires two separate products, and the court did not see two. It quoted its earlier ruling: “If AI Overviews is a separate product from Google Search, then so is potentially every other type of search result Google delivers.” [1] Then it turned the publishers’ own evidence against them. If users wanted search only to travel to other websites, they would scroll past the summary. The complaints alleged the opposite, that users stop at the answer. Those allegations, the court wrote, suggest “a general demand for information and consumer indifference as to how they obtain it.” [1]
In other words, the better AI Overviews are at keeping the reader, the weaker the argument that they are a different product from search. The court also noted that other search engines, such as DuckDuckGo, put a generated summary at the top of their results too. [1]
| Claim | Brought by | Outcome | The court’s reason |
|---|---|---|---|
| Reciprocal dealing | Both | Dismissed | No agreement was pleaded [1] |
| Tying AI Overviews to search | Penske only | Dismissed | Not plausibly separate products [1] |
| Monopoly maintenance in general search | Both | Dismissed | No antitrust standing; the injury is too secondary and indirect [1] |
| Attempted monopolization and leveraging | Both | Dismissed | No plausible market definition [1] |
| Unjust enrichment | Both | Dismissed, not decided | The court declined jurisdiction over state-law claims [1] |
What the court said it could not do
The last row of that table matters. The unjust enrichment claims were not rejected. With every federal claim gone, the court chose not to hear the state-law ones, which is the usual course. [1] The order in the Penske case grants the motion “without prejudice” and calls itself a final, appealable order. [2] The fight can continue, on appeal or elsewhere. What is settled for now is narrower: on these complaints, antitrust law does not turn the old exchange of content for clicks into an obligation.
The opinion does not address copyright. These were antitrust complaints, and the ruling should not be read as a judgment on whether using a page in a generated answer is lawful in every respect.
The judge was direct about the limits of his own role. The court, he wrote, “does not treat Plaintiffs’ alleged harms lightly,” and is not unsympathetic to “journalists, educators, and other online creators whose content Google takes and repurposes without compensation.” [1] But where a case exposes a gap in the antitrust laws, the opinion says, quoting an earlier decision, “that is an issue for Congress [or regulators] to consider.” [1]
Compare that with the other Google case decided last month, where a different court ordered Google to open its ad exchange to rivals. There, the government had proved a monopoly was maintained unlawfully. Here, private plaintiffs could not get past the pleading stage.
The control Google already documents
The court’s view that AI answers are part of search matches what Google tells site owners. “AI is built into Search and integral to how Search functions,” its documentation says, “which is why robots.txt directives for Googlebot is the control for site owners to manage access to how their sites are crawled for Search.” [5]
The finer controls are the snippet ones: nosnippet, data-nosnippet, max-snippet, and noindex. [5] They are blunt. The complaints themselves, as the opinion records, said that opting out of snippets would “prevent snippets from being shown as previews in search results.” [1] Google-Extended is a separate setting, and Google describes it as covering training and grounding in some of its other systems, not Search. [5]
So the practical position has not moved. A site can be in Google Search with its AI features, or limit how it appears everywhere at once. We laid out what that costs in clicks in our analysis of AI Overviews traffic.
A price list in HTTP
Cloudflare’s announcements are an attempt to supply what the court said was missing: terms, offered and accepted.
Pay Per Use, now in beta, works like this. An AI company identifies its crawler, defines the use it will pay for, and sets a price. A publisher sees the offer in its dashboard and accepts or declines. The buyer reports each use through an API, and Cloudflare bills the buyer and pays the publisher monthly. [3] Cloudflare contrasts it with Pay Per Crawl, the 2025 product that charges a crawler for access through an HTTP 402 response. [3][6] The idea is that a page may be crawled once and used a thousand times, so the use is the thing worth pricing. [4]
The weak point is stated in the post itself: “Usage is self-reported: the program terms require complete reporting, and Cloudflare checks that each reported use maps to an enrolled publisher.” [3] The publisher sees what the buyer says it used.
The second product, Monetization Gateway, entered closed beta the same day. It lets a site owner charge agents per request for APIs, tools, datasets, or pages, with payment handled inside the request using the HTTP 402 status code and settled in USDC on the Base blockchain through Coinbase’s x402 facilitator. Cloudflare calls the simplest version a “paywall for agents.” [4]
Both products need a buyer who agrees to pay. That is their limit, and it is the same limit the lawsuits ran into. The AI companies most likely to sign up are the ones that need access they do not already have. Google crawls for its AI answers with the same Googlebot that sites depend on for search, and a toll in front of Googlebot is a toll in front of your own rankings. [5]
Google does appear to pay some publishers on its own terms. PPC Land, summarizing a report in The Information that we could not read at its source, says about 100 sites are paid through a pilot, with several small and midsize sites receiving less than $1,000 over several months and one early participant earning more than $1 million a year. [7] Treat those as reported figures. If they are right, they describe a program where Google sets the price and explains little, which is the opposite of a negotiated deal.
What to do with this
Plan on the assumption that clicks from Google’s AI answers will stay uncompensated and inseparable from search for the foreseeable future. No court order is coming soon to change that, and the judge who wrote this one pointed at Congress.
Decide crawler access bot by bot, by what each one is for. Training crawlers, search crawlers, and user-triggered fetchers are different decisions, and only some of them have a buyer on the other side.
If your site runs on Cloudflare, look at the Pay Per Use offers when they appear, and read the terms on training. Cloudflare says each program’s terms define what the AI company may do with the content. [3] Treat the earnings report as a statement from the buyer, because that is what it is.
And stop describing search traffic as something you are owed. The court’s sentence is uncomfortable because it is accurate about how the arrangement always worked. The traffic was a side effect of a product decision. Product decisions change.
Sources
- US District Court for the District of Columbia · Memorandum Opinion, Penske Media Corp. v. Google LLC, No. 25-cv-3192, and Chegg, Inc. v. Google LLC, No. 25-cv-543Primary. Filed September 30, 2026; 41 pages. Via CourtListener.
- US District Court for the District of Columbia · Order, Penske Media Corp. v. Google LLCPrimary. Filed September 30, 2026. Via CourtListener.
- Cloudflare · Pay Per Use: when AI uses your work, you should get paidVendor announcement, September 30, 2026.
- Cloudflare · Monetization Gateway beta: charge AI agents for consumption with HTTP 402Vendor announcement, September 30, 2026.
- Google Search Central · AI features and your websitePrimary.
- Cloudflare · Introducing pay per crawl: Enabling content owners to charge AI crawlers for accessVendor announcement, July 1, 2025.
- PPC Land · Google pays about 100 publishers for AI answers, some under $1,000Trade press summary of a report in The Information, which we could not read at source. Figures not confirmed by Google.
Frequently asked questions
What did the court decide?
On September 30, 2026, Judge Amit Mehta of the US District Court for the District of Columbia granted Google’s motions to dismiss the amended complaints in Chegg v. Google and Penske Media v. Google. One 41-page memorandum opinion covers both cases. The order in the Penske case grants the motion without prejudice and states that it is a final, appealable order.
Why did the main claim fail?
The publishers alleged reciprocal dealing: that Google conditions search referral traffic on receiving their content for free. The court held that they had not pleaded any agreement. They did not allege that Google offered terms they accepted, that anything was negotiated, or that either side made a commitment. An expectation of traffic, the court wrote, is not an agreement.
Did the court say AI Overviews are legal?
It ruled on the claims in front of it, which were antitrust claims and a state-law unjust enrichment claim. It found those antitrust claims were not adequately pleaded. The opinion does not address copyright, and it expressly declined to rule on unjust enrichment.
What happened to the unjust enrichment claims?
They were not decided on the merits. After dismissing every federal claim, the court declined to exercise supplemental jurisdiction over the state-law unjust enrichment claims and dismissed them on that basis.
Can a site opt out of AI Overviews and stay in Google Search?
Not with a dedicated switch. Google’s documentation says AI is built into Search and that robots.txt directives for Googlebot are the control for how sites are crawled for Search. To limit what is shown, it points to nosnippet, data-nosnippet, max-snippet, and noindex, which also affect ordinary results. Google-Extended covers training and grounding in some of Google’s other systems.
What is Cloudflare Pay Per Use?
A beta program announced on September 30, 2026. An AI company defines a use it will pay for and sets a price, a publisher chooses whether to accept, the buyer reports each use through an API, and Cloudflare bills the buyer and pays the publisher monthly. Cloudflare states that usage is self-reported.
Does Google pay publishers for AI answers?
There are reports that it pays some. PPC Land, summarizing a report in The Information that we could not read at its source, says about 100 sites are paid through a pilot, with several small and midsize sites receiving less than $1,000 over several months and one early participant earning more than $1 million a year. Google has not published those figures.
Mikołaj Salecki
Writes about media, tech, and AI business for people who actually run digital. Former agency lead. Skeptic of frameworks that read better than they perform.
