Judge Dismisses AI Overviews Suits: What Publishers Can Do

On September 30, 2026, Judge Amit P. Mehta of the US District Court for the District of Columbia dismissed the antitrust suits that Chegg and Penske Media Corporation brought against Google over AI Overviews. Both publishers argued that Google forces them to hand over content for free in return for search traffic. The court found no agreement of that kind, and every federal claim failed.

Editorial note: Prepared October 3 as an October 2, 2026 explainer from the court’s memorandum opinion in Chegg, Inc. v. Google LLC, No. 25-cv-00543, and Penske Media Corp. v. Google LLC, No. 25-cv-03192, filed September 30, 2026, and from Google’s Search Central documentation. This is a summary for publishers, not legal advice.

Chegg, an education technology company, sued in February 2025. Penske Media, which owns Rolling Stone, Billboard, Variety, The Hollywood Reporter and Deadline, sued in September 2025 with several of its publications as co-plaintiffs. The court heard Google’s motions to dismiss together and decided them in one opinion. The table sets out each claim and the court’s reason for rejecting it.

The market-definition point is worth a second look. Penske defined an online publishing market covering nearly every piece of text published online. The court asked why a blog post, a legal brief, a short story and a news article would be substitutes for one another, and found the market too broad to be plausible. Chegg’s narrower market for online educational publishing failed for a different reason. Chegg defined it by curation and verification, but its own complaint described Google’s models scraping and reusing content with neither, so Google did not plausibly compete in that market.

Both complaints rested on a bargain. Publishers let Google crawl their pages; Google sends readers back. The publishers said Google now makes them supply three kinds of content for free as a condition of that traffic: content for snippets, content to train its models, and content its AI features retrieve to answer questions. AI Overviews, they said, breaks the bargain by answering the question without the click.

The court’s answer was that there was never a bargain in the legal sense. The complaints described no offer, no negotiated terms and no commitment on either side, only the publishers’ hope that letting Google index their pages would bring traffic.

This was not the first time. In March 2026, the same judge dismissed similar claims in Helena World Chronicle v. Google, and the new opinion leans on that ruling for its points on tying and standing. Two consecutive decisions on the same theory make it a hard route for the next publisher who tries it in this court.

The opinion closes less than the headlines suggest. Three things remain.

• An appeal. The opinion comes with a final, appealable order, so either publisher can take it to the D.C. Circuit.
• The state-law claims. The court did not rule on unjust enrichment. It declined to hear those claims once the federal ones were gone, which is different from finding them without merit.
• Other legal theories. These were antitrust suits. The opinion does not address copyright or licensing, which are separate questions decided under separate law.

The judge was also explicit that the outcome is a limit of the law, not an approval of the practice. The opinion says the court does “not treat Plaintiffs’ alleged harms lightly” and acknowledges the effect on journalists, educators and other creators whose content Google uses without payment. It then says that if antitrust law does not reach the problem, that is for Congress or regulators to fix.

The ruling changes nothing about the technical options, and the complaints themselves explained why publishers dislike them. Google’s documentation on AI features lists four ways to limit what Search shows from a page: the nosnippet and max-snippet rules, the data-nosnippet attribute on part of a page, and noindex. All four apply to Search as a whole, not to AI Overviews alone.

The last row is the one most often misunderstood. Google’s crawler documentation says Google-Extended does not affect a site’s inclusion in Google Search and is not a ranking signal. Blocking it does not take a page out of AI Overviews. Measurement is limited too: Google says Search Console includes traffic from AI features in the overall web search totals, and its documentation describes no separate AI Overviews figure.

For most publishers, the practical response is to measure first and control second. Start with the pages that matter. Find the queries where traffic fell while impressions held, which is the pattern our traffic drop diagnosis guide walks through, and test data-nosnippet on a small set of those pages before applying anything site-wide.

Separate the AI crawlers you can charge or block from the one you cannot. Google’s Search crawler is tied to Search traffic. Other AI crawlers are not, and the options for charging them are growing, as we covered in Cloudflare’s pay-per-use and payment gateway options. And keep publishing what an AI summary cannot replace: original reporting, data and tools. Google’s own October 1 guidance on AI content points the same way.

Where search is a large share of revenue, this is worth a structured review rather than a one-off change. Our agentic SEO work covers the measurement, the controls and the content plan together.