Judge Throws Out Google’s Scraping Suit Against SerpApi a Second Time

BySarah Fraser
A laptop screen displaying a Google search results page, with blue links and snippets, in a dimly lit room.

Google built its search engine by copying other people’s websites without asking. A federal judge has now twice thrown out its lawsuit to stop a small company from copying Google. SerpApi sends automated searches to Google disguised as ordinary users, then sells the results to customers including Nvidia, Uber, Adobe and Perplexity. It has about 42 staff and revenue of “a few million dollars”. Google’s search business took in $63.35 billion in the third quarter of 2026.

On July 20, Chief Judge Yvonne Gonzalez Rogers threw out Google’s claim over ordinary search results for good. On November 9 she dismissed what remained, a narrower claim over content Google licenses from others. Google has until November 30 to try a third time.

Google’s complaint says SerpApi uses Google’s computing power without paying for it and undercuts the deals Google signs to license content. In January 2025 Google launched SearchGuard, an anti-bot system it says cost tens of thousands of person-hours and millions of dollars. SerpApi got around it. Its founder described the method as:

“creating fake browsers using a multitude of IP addresses that Google sees as normal users.”

According to the complaint, SerpApi would solve SearchGuard’s challenge once with a “legitimate” request and then share the resulting pass with machines around the world. Google says the result is hundreds of millions of fake queries a day. Google handles about 13.7 billion searches a day, which would put SerpApi at roughly 1.5% to 6.5% of all Google traffic.

The licensing deals are where the money is. Google’s deal with Reddit was reported at about $60 million a year, and it tells Google “not to enable third parties to extract and independently commercialize the licensed content.” General counsel Halimah DeLaine Prado said SerpApi “deceptively takes content that Google licenses from others.” SerpApi says it provides only information that anyone can see in a standard browser, and that Google is trying to limit competition from AI and web services.

Those competitors have few other options. Google’s own Custom Search API is capped at 10,000 queries a day, is closed to new customers and shuts on January 1, 2027. Its full-web search API is available only to partners under agreement. SerpApi charges about $9 to $25 per 1,000 searches, compared with Google’s $5. Its customers pay more because they need volume that Google will not sell them.

Google sued on December 19, 2025 under Section 1201 of the Digital Millennium Copyright Act, which makes it illegal to break a digital lock on a copyrighted work. The law was written against DVD and software piracy, and companies have since used it against printer ink refills and garage door openers. Damages run from $200 to $2,500 per act of circumvention. SerpApi calculates that the $200 minimum, applied to Google’s own count, comes to $7.06 trillion.

The catch is that the lock has to protect a copyrighted work, and links, snippets and facts are not copyrighted works. On July 20 the court ruled:

“SearchGuard cannot be said to effectively control access to a work protected under the Copyright Act.”

That claim cannot be refiled. The claim that survived covered licensed material, such as images in Knowledge Panels, the information boxes Google shows beside results. On August 10 Google refiled and named four sources of permission: the Reddit licence, an unnamed “prominent licensing partner,” an unnamed “major content provider,” and its own Privacy Policy. On November 9 the court rejected all four. It called the allegations about the prominent partner “bare and conclusory.”

Google did win some arguments, and the judge has not ruled that SerpApi’s scraping is lawful. Across the two orders the court decided:

Decided for Google Decided for SerpApi
Spoofed browsers, rotating IP addresses and CAPTCHA-solving count as circumvention Ordinary search results are not copyrighted works (final)
Google has standing, the right to bring the suit The Reddit licence doesn’t show owners authorised SearchGuard
Offering a circumvention service is enough, with no sale needed Claims about the “prominent licensing partner” are “bare and conclusory”
Permission from owners need not be a written contract Google’s Privacy Policy is not permission from anyone

Google spokesperson José Castañeda said the company was “pleased to see that the Court rejected nearly all of SerpApi’s legal arguments.” Techdirt’s Mike Masnick opposes the lawsuit, yet he still calls SerpApi’s behaviour “sketchy.” Google has alleged that SerpApi breaks the lock, and the court found that allegation sufficient. What Google cannot show is that the lock protects anything Google owns.

Google’s crawling, and everyone else’s

Meredith Rose, a DMCA specialist at the nonprofit Public Knowledge who is not involved in the case, told Ars Technica:

“They have talked themselves into a little bit of a corner.”

To win, Google has to plead that its results pages are full of other people’s copyrighted work, collected and displayed in bulk. That also describes Google’s AI Overviews. Every rights holder whose work fills those pages would then have a Google filing that agrees with them.

SerpApi relies on exactly this point. In its February motion to dismiss, a request to end the case before any evidence is heard, it called Google “the largest scraper in the world” and argued that “Mimicry is not a violation.” Google’s reply is that it follows “industry-standard crawling protocols” and honours robots.txt, the file sites use to ask crawlers to stay away. But in Ziff Davis v. OpenAI, Judge Sidney Stein of the Southern District of New York ruled in December 2025 that robots.txt is a “keep off the grass” sign, not a lock. Publishers’ fences are signs Google can choose to honour; Google’s fence is a lock.

Publishers who tried to use Google’s crawling against it have lost. Penske Media sued over AI Overviews after its affiliate revenue fell by more than a third. Judge Amit Mehta dismissed the case, ruling that publishers had “voluntarily acceded” to free crawling and that “an expectation is not an agreement.”

Publishers’ fences are signs Google can choose to honour; Google’s fence is a lock.

Taken together, the rulings work like this. Publishers who left their pages open gave Google consent by doing so. Google locked its own pages and still cannot use this law to stop the copying, because what sits behind the lock belongs to those same publishers.

Illustration of a web crawler copying pages from publisher websites into a search engine, while a padlock guards the search engine's own results page.

Where the anti-scraping law does work

Reddit’s case involves the same scraper, the same lock and the same Google results pages, and Reddit got past dismissal. It sued SerpApi under Section 1201 over Reddit content taken from Google results, and Judge Paul Engelmayer let the claims go ahead in late July, because Reddit alleged access to content “authored by Reddit and its users.” Google wrote none of the content in its results, and its case did not survive.

YouTube creators are winning the same way. Ted Entertainment, the company behind h3h3 Productions, and golfer Matt Fisher sued Apple for scraping their videos to train AI. Apple argued the videos were public. On October 9 Judge Richard Seeborg refused to dismiss the case and held that a lock need only give “some degree of control over access.” The creators allege Apple defeated YouTube’s download protection and switched IP addresses, the same methods SerpApi used. It is the third creator case to survive a motion to dismiss, after suits against ByteDance and Snap.

In each of those cases, the plaintiff owns or wrote what was behind the lock. Google is losing because it does not.

To refile by November 30, Google must plead facts showing which owners authorised SearchGuard and how. If it misses the deadline, the case is dismissed with prejudice, meaning it is thrown out for good. Google makes that decision while its search business is slowing. Search revenue grew 12% year on year in Q3 2026, down from 17% the quarter before, and rose just $79 million from Q2.

Winning would mean telling a federal court that Google’s results pages are built from other people’s property. Publishers have been telling Google that for twenty years.

Further Reading