Google filed an amended complaint against SerpApi on August 10, 2026, rescuing a Digital Millennium Copyright Act case that Judge Yvonne Gonzalez Rogers had dismissed in full on July 20. The rescue carries a price that runs well past this docket. To keep the case alive, Google pleaded its confidential licensing contracts, one with Reddit and two with partners it declines to name, as the factual core of its claims, placing that paper on the road to discovery. The arguments doing the rescuing mirror the ones publishers and regulators already press against Google itself, through federal antitrust suits, a European enforcement docket and binding orders on two continents to open search data outward. A lawsuit built to stop the scraping of search results is turning into a map of Google's own exposure.

The amended complaint keeps the same two counts under Section 1201 of the DMCA, the statute's anti-circumvention provisions. Count one accuses SerpApi of circumventing SearchGuard, the JavaScript challenge system Google deployed in January 2025 to separate human visitors from automated software. Count two accuses the Austin, Texas company of trafficking in circumvention services. What changed is the foundation underneath those counts. Where the December 19, 2025 original leaned on Google's investment in search generally, the new filing stakes everything on authorization from copyright owners, because the court told Google in plain terms that nothing else survives.

The order that redrew the case

The July 20 ruling, covered by PPC Land at the time, split Google's claims in two. For search results pages containing no copyrighted content, Judge Gonzalez Rogers dismissed without leave to amend, writing that "amendment would be futile" because the claims fail as a matter of law. The DMCA protects copyrighted works, and a compilation of publicly available links, in the court's reading, is not one. That portion of the case is dead.

For results that do contain a copyrighted component, such as licensed photographs inside Knowledge Panels, the court left a path open, but a demanding one. Under 17 U.S.C. § 1201(a)(3)(B), a technological measure counts as an effective access control only if it operates "with the authority of the copyright owner." According to the order, Google alleged no facts raising the inference that the photographers, music publishers, platforms and other rights holders whose material appears in search results ever authorized SearchGuard to guard it. Licenses to display content, the judge held, do not by themselves imply licenses to lock it up.

Notably, the order rejected several of SerpApi's broader arguments. Standing under the DMCA is not limited to copyright owners, the court found, because Section 1203(a) permits suit by any person injured by a violation. The court also held that Google adequately alleged circumvention itself, pointing to allegations that SerpApi masks automated queries as human traffic and syndicates solved challenges to machines that never solved them. Those rulings matter now. They mean the amended complaint faces one central question rather than five: did the copyright owners say yes?

What changed on August 10

The 15-page amended complaint arrived on the final day of the 21-day window the dismissal order allowed. The answer Google supplies runs through four new categories of authorization, laid out in paragraphs 30 through 33 of the amended filing. According to the complaint, Google holds an agreement with a prominent licensing partner, unnamed in the document, covering millions of copyrighted works, with contract versions dating back to 2017. That agreement obligates Google to use "commercially reasonable efforts" to safeguard the licensed content against unauthorized third-party access. A second major content provider, also unnamed, requires Google to ensure its licensed material "will not be available for download by third parties," according to the filing.

The third pillar is Reddit, and it is the one Google names. The amended complaint states that Reddit's licensing agreement directs Google "not to enable third parties to extract and independently commercialize the licensed content." Google also quotes Reddit's own amended complaint in a parallel New York case, in which the platform stated that it "has always authorized its partners, including Google, to protect" licensed content from misuse. According to Google's filing, Reddit went further when it suspected scrapers were harvesting its posts from search results and specifically asked Google to deploy technical measures against the practice. The fourth category reaches into Google's Privacy Policy, which promises users protection of uploaded content, such as photos and reviews, from "unauthorized access."

The restructuring extends beyond the authorization paragraphs. References to Google Shopping and Google Maps as separate services, which the court flagged as irrelevant to a case about Search results, have been reworked; the amended complaint now describes merchant-supplied product imagery, music lyrics and user-generated content as material appearing within Search results themselves. PPC Land documented the revived claims when the filing became public this week.

Discovery now points at Google's own contracts

The nearest-term consequence of the strategy is procedural, and the July 20 order previewed it. Judge Gonzalez Rogers stayed discovery until the amended complaint is filed and any renewed motion to dismiss is resolved, denying along the way Google's request for an order compelling SerpApi to answer its first set of document requests and interrogatories within ten days. The order added a sentence with a second edge: the authorization information Google needed for its amendment already sat, in the court's estimation, within its "possession, custody, or control." That phrase tracks the language of Federal Rule of Civil Procedure 34, which defines the material a party must hand over in discovery. The court has, in effect, already located the relevant documents. They are inside Google.

By pleading the contracts, Google converted them from confidential commercial paper into the factual core of its case. Authorization from copyright owners is now an element Google must prove, and material relevant to a party's claims is discoverable under Rule 26 once the stay lifts. The two agreements the amended complaint declines to name, the prominent licensing partner holding millions of works and the provider requiring that content not be available for third-party download, can remain anonymous in a public pleading but not in discovery, where SerpApi can demand their identities, the complete agreements, drafts and the negotiation record behind the clauses Google quotes. Protective orders and sealing, routine in the Northern District of California, restrict what the public sees; they do not keep documents from opposing counsel and experts. Pleaded communications travel the same road. The complaint states that Reddit specifically asked Google to deploy technical measures against scrapers, and an alleged conversation is a producible one.

The scope extends past the contracts themselves. The court's division between results with and without copyrighted content makes SearchGuard's actual design and targeting relevant, terrain SerpApi has already mapped with its argument that the system operates at the domain level rather than around protected works. Timing questions follow: the lead agreement exists in versions that, according to the complaint, "date back to 2017," while SearchGuard launched in January 2025, inviting examination of which works were covered, and authorized for protection, at the moment of each alleged circumvention. A damages election adds a further layer. Section 1203 lets Google choose statutory damages, but pursuing actual damages instead would put the economics of the licenses and the cost of processing SerpApi's queries into the record.

Litigation has pried open Google's paper before. The search antitrust case placed on the public record that Google paid an estimated $20 billion to Apple in 2022 for default placement on Safari, a figure guarded for years before trial exposed it. The licensing terms behind Knowledge Panels would enter the same machinery, and Reddit sits in an unusual position within it: Google's licensor, a plaintiff against SerpApi in New York, and a counterparty negotiating a renewal while its existing terms head toward production in an Oakland docket. Subpoena traffic between the cases already flows in one direction, since Reddit obtained data from Google quantifying SerpApi's results-page accesses at more than 1.8 billion pages containing Reddit data across two weeks of July 2025.

The timing turns on SerpApi's next move. A response to the amended complaint is due within 14 days of service under the federal rules. Another motion to dismiss keeps the stay alive, though the order bars arguments that could have been raised the first time. An answer would end the stay at once. Either way, the moment this case holds claims that survive, the first documents on the table belong to the plaintiff.

The exposure Google creates for itself

The exposure extends beyond this docket, and it is not hypothetical. Publishers are already suing Google over uncompensated content extraction, and the SerpApi filings hand them material. Penske Media filed a 101-page federal antitrust lawsuit on September 12, 2025, alleging Google uses its search monopoly to coerce publishers into supplying content for AI training, AI Overviews and AI Mode without compensation, and seeking treble damages under the Sherman Act. The Atlantic followed with a federal suit in January 2026 and Dotdash Meredith sued in August 2025, filings catalogued in coverage of Penske's February 2026 opposition briefing. The European Commission opened a formal investigation on December 9, 2025 into whether Google used publisher and YouTube content for AI purposes without appropriate compensation or a viable opt-out, a probe that arrived alongside research showing AI Overviews cut publisher clicks by 39.8 percent. In Germany, the Munich Regional Court held Google directly liable on May 28, 2026 for false claims its AI Overviews generated about two publishing companies, backed by fines of up to EUR 250,000 per violation, stripping a liability shield the company had held for two decades.

Into that environment, Google has now filed pleadings stating that automated scraping at scale imposes a deadweight loss on the scraped service, that taking content without permission or compensation undermines investment in it, that redistribution denies rights holders payment for their works, and that robots.txt instructions carry legal weight. Each proposition mirrors, nearly clause for clause, what publishers allege about Google's own crawlers, and opposing litigants read each other's filings. SerpApi's general counsel made the connection in January 2026, arguing that Google restricts access to the same kind of publicly displayed information it takes freely from everyone else.

The DMCA mirror is sharper still, though it points at scrapers generally rather than at Google directly. Under the July 20 construction, a Section 1201 plaintiff needs a copyrighted work, a technological measure, owner authorization and circumvention. A publisher enforcing a bot wall over its own staff-written content clears the first three automatically; it is the owner, so the authorization gap that sank Google's original complaint does not exist for it. What is missing against Google specifically is the fourth element. Googlebot identifies itself, and Google states its crawlers honor robots.txt, so publisher grievances center not on bypassed controls but on a forced bundle: blocking AI use means disappearing from search results, the coerced choice at the heart of Penske's complaint. That grievance sounds in antitrust and regulation rather than copyright, which is where publishers are pressing it, and not always successfully. A Washington, D.C. court dismissed the Helena World Chronicle publishers' antitrust case in April 2026 for lack of standing. Judge Amit Mehta, who decided that dismissal, hears Penske's case as well.

The antitrust collision around the case

Mehta's other docket frames the sharpest structural conflict. His August 5, 2024 ruling found Google illegally maintained monopolies in general search services and search text advertising, and the final judgment entered December 5, 2025 imposed six-year behavioral remedies that include mandatory data sharing and search syndication services for qualified competitors, with the bulk of the judgment taking effect on February 3, 2026. Google asked the D.C. Circuit for full reversal in a 100-page opening brief on May 22, 2026. The result is a company under court order to license its search results outward on regulated terms while asking a different federal court to price unlicensed access to the same results at $200 to $2,500 per query. Between those channels sits a gap with commercial consequences: the syndication remedy contemplates rival search engines and AI answer engines as recipients, not data resellers, so a service like SerpApi would hold neither mandated access under the antitrust judgment nor tolerated access under a successful DMCA theory.

The monopoly finding also changes how Google's access restrictions read. Product decisions by an ordinary firm become Section 2 material when an adjudicated monopolist makes them, and the sequence since the liability ruling is a matter of record: the n=100 results parameter disappeared on September 14, 2025, forcing roughly tenfold cost increases for SEO tools, the Light Fast API workaround was restricted to three organic results in early October 2025, and SearchGuard operates across the google.com domain regardless of whether a given page contains licensed content, the characterization from SerpApi's motion that the court partly credited. SerpApi has framed the lawsuit itself in antitrust vocabulary, with its general counsel arguing in January 2026 that the litigation works to raise rivals' costs and limit competition. Whether or not that framing ever matures into a counterclaim, it positions any Google victory as evidence in the continuing fight over how much control the company retains over access to results.

Brussels runs the same collision harder. On July 16, 2026, the European Commission adopted binding Digital Markets Act decisions requiring Google to share anonymised search data, covering query, click, ranking and view data, with rival engines including AI chatbots, after finding that Google's first compliance offer had stripped out between 90 and 100 percent of unique search queries. The Court of Justice confirmed the EUR 4.1 billion Android fine on July 2, 2026, and the Commission's December 2025 probe into Google's AI content practices remains open. A company under binding orders on two continents to open elements of search outward is simultaneously asking a California court to hold that taking its results without a license violates federal copyright law.

The appellate fight runs in both directions. A bipartisan amicus brief filed August 4, 2026 urged the D.C. Circuit toward tougher remedies, citing concentration of 88 percent of AI model API revenue in three firms. If the appeals court vacates the syndication remedy while the DMCA theory survives, Google would hold both ends of the transaction: no compelled licensing out, statutory damages against unlicensed taking in. If the remedies stand and the DMCA theory fails a second time, the flow reverses. The Oakland pleadings, which describe Google's results as the product of billions of dollars of investment and licensed value, supply exactly the kind of characterization that remedy advocates deploy when arguing access to those results is worth mandating.

A theory that cuts both ways

The question hanging over the amended complaint is whether Google has walked into a structure it cannot easily exit. The court's construction makes the DMCA shield coextensive with Google's licensing footprint. Organic results assembled from the open web enjoy no protection under this theory; the ruling on that point is final at the district level. Protection attaches only where content is licensed and where the license, or the rights holder directly, authorizes technical enforcement. Every fact Google added on August 10 is a contract term. The case now rises or falls on paper Google signed with its partners, which gives those partners a new kind of weight.

Reddit illustrates the shift. The platform signed a licensing agreement with Google in February 2024 reported at roughly $60 million per year, supplying training material for Gemini and effective exclusivity over indexing recent Reddit content. That deal entered renewal talks in July 2026, at a moment when several large publishers were weighing whether to cut Google's crawler off entirely. Reddit's contract language is now load-bearing evidence in Google's federal case. A partner whose terms sustain the plaintiff's standing sits in a stronger negotiating position than one supplying training data alone, and each future licensing agreement that includes an authorization clause widens the zone SearchGuard can lawfully defend.

The same architecture is being assembled on the other side of the crawl economy, aimed at companies that fetch content at scale, a category that includes Google. Cloudflare launched pay-per-crawl on July 1, 2025, letting publishers charge crawlers through HTTP 402 payment responses, then shifted toward per-answer publisher payments on July 1, 2026 while announcing that training and agent crawlers will be blocked by default on ad-supported pages of new domains from September 15, 2026. A precedent holding that an anti-bot challenge, backed by copyright-owner authorization, qualifies as a Section 1201 access control would be available to any publisher operating such systems against any scraper. Google has resisted the commercial half of that logic in public: when asked in a December 15, 2025 podcast interview about standardized licensing frameworks for publishers, Google search executive Nick Fox rejected the idea, as PPC Land reported in its analysis of the company's dual role as plaintiff and crawler.

A third pressure runs through antitrust court. The final judgment in the Department of Justice search case, entered December 5, 2025, requires Google to provide syndication services and data access to qualified competitors; Google appealed and asked to pause those remedies on January 16, 2026. One federal proceeding compels licensed distribution of Google's results. Another, initiated by Google, seeks to make unlicensed taking of those results a federal violation priced at up to $2,500 per query. Both push toward the same endpoint from opposite directions: search results as a product that moves only under license.

There is a limit to the trap. Section 1201 protects access controls over copyrighted works; it creates no general right to charge for crawling, and the court has already ruled that the broad version of Google's theory fails as a matter of law. What the amended complaint can entrench, if it survives, is narrower and still significant: a regime in which the enforceability of a bot wall depends on the licensing contracts behind it, giving every rights holder in the chain, from Reddit to unnamed image libraries, a legal and commercial stake in how, and at what price, automated access happens.

The numbers that frame the stakes

The scale allegations remain the case's gravitational center. Google estimates that SerpApi sends hundreds of millions of artificial search requests to Google each day, a volume that, according to the amended complaint, has grown by as much as 25,000 percent over the past two years. Each circumvention carries statutory damages of between $200 and $2,500 under Section 1201. Multiplied across what Google describes as billions of separate occasions, the arithmetic produces liability figures the complaint itself acknowledges SerpApi cannot pay: the defendant reportedly earns a few million dollars in annual revenue, while facing exposure "orders of magnitude higher and growing," according to the filing. Google presents that gap as the argument for an injunction rather than a damages award.

SearchGuard, for its part, represents tens of thousands of person hours and millions of dollars of investment, according to the complaint. SerpApi, founded in 2017 after chief executive Julien Khaleghy concluded that "scraping images from Google was an intensive process," responded to the system's January 2025 launch within weeks. The amended complaint quotes Khaleghy describing the workaround as "creating fake browsers using a multitude of IP addresses that Google sees as normal users," and cites a SerpApi blog post claiming the company had "already pre-solved Google's JavaScript challenge" when SearchGuard arrived.

SerpApi's answer and the Reddit front

SerpApi has signaled no retreat. "We remain confident in the strength of SerpApi's position and look forward to responding," company counsel said on August 11, 2026, according to a MediaPost report on the amended filing. The company's public posture has been consistent since general counsel Chad Anson published a point-by-point rebuttal on January 23, 2026, arguing the service retrieves only what any browser can see. Its February 20, 2026 motion to dismiss contended that SearchGuard operates at the domain level, blocking bots regardless of whether any given page contains licensed material, an argument the court partly accepted and partly rejected.

The parallel litigation is moving in the opposite direction. Reddit named Perplexity AI alongside SerpApi, Oxylabs and AWMProxy in a lawsuit filed October 22, 2025 in the Southern District of New York, alleging the defendants bypassed SearchGuard to harvest Reddit posts from results pages. Subpoena data cited in that case, reported by PPC Land in March, put SerpApi's accesses at 784,006,306 results pages containing Reddit data between July 1 and July 6, 2025, and 1,057,744,585 between July 7 and July 13. In late July 2026, Judge Paul Engelmayer ruled that Reddit could proceed with its claims, according to MediaPost. Two federal courts examining the same conduct have, for the moment, reached opposite preliminary conclusions, with the distinction turning on how each complaint pleaded ownership and authorization.

What separates Reddit's treatment from everyone else's

Reddit's position rests on four elements: a licensing contract reported at roughly $60 million per year, effective exclusivity achieved by blocking other search crawlers through robots.txt, contract clauses obligating Google to protect the licensed content, and now a federal plaintiff litigating partly on Reddit's paper. None of that came from a standard. It was negotiated by a platform whose conversational archive Google wanted badly enough for Gemini training, at a moment when Reddit could credibly withhold it. Google's Nick Fox rejected standardized licensing frameworks for publishers in the December 2025 interview cited above, which leaves individual negotiation, and a site with a fraction of Reddit's corpus negotiates from a position Google has little commercial reason to accommodate.

The legal half of Reddit's treatment travels more easily than the commercial half. The elements Judge Gonzalez Rogers spelled out describe a case that a publisher of its own content can plead more cleanly than Google could, because ownership eliminates the authorization question entirely. Statutory damages of $200 to $2,500 per act scale with the scraper's query volume, not with the plaintiff's size. The late July ruling from Judge Engelmayer lets Reddit pursue scrapers for content harvested from Google's pages rather than from Reddit's own servers, extending an owner's reach beyond its own infrastructure. SerpApi's defense in that case also marks the soft spot: platforms built on user-generated content hold non-exclusive licenses from their users, a weaker footing than a publisher that owns its articles outright.

Infrastructure has meanwhile begun substituting for bargaining power. Any Cloudflare customer can price crawler access through pay-per-crawl, and from September 15, 2026 new ad-supported domains on that network receive default blocking of training and agent crawlers, a default refusal that previously required Reddit-scale weight to impose. The IAB Tech Lab formed its Content Monetization Protocols working group on August 20, 2025 with roughly 80 executives to standardize Cost per Crawl pricing and bot management. Time now blocks all AI bots by default and routes whitelisted operators through TollBit's marketplace to stripped-down markdown versions of its pages. Four midsize media brands began licensing content under paid, attributed terms through the SAIL framework in July 2026, built on NVIDIA infrastructure.

What the mechanisms have not closed is the gap in price. Reddit's data licensing revenue reached $36 million in a single quarter, while one publishing company earned a total of $174 from AI crawler payments over an extended period, a figure disclosed on November 20, 2025 and documented in PPC Land's analysis of the crawl economy. The tools that once required Reddit's scale now come bundled with a CDN account or a trade-body membership. The prices do not.

Small publishers absorb the losses on every front

The prices are one gap. The traffic is another. Under the July 20 construction, the legal wall Google is building protects only the content it licenses; organic listings, the layer where independent sites live, were dismissed from the theory with prejudice. The same open-web material feeds the AI features that cut clicks to its sources. A randomized study found AI Overviews reduce outbound clicks to publishers by 39.8 percent, and the IAB Tech Lab framework put average traffic losses from AI search summaries at 20 to 60 percent, with niche sites losing as much as 90 percent, against an estimated $2 billion in annual ad revenue. The content that trains and grounds those answers carries no license fee at this end of the market, and under Google's own legal theory, no protection either.

The casualties have names. Overfishing.org, a conservation site that ran for 21 years, shut down entirely after its traffic was absorbed by AI Overviews, while All About Berlin, an eight-year-old independent expat guide, lost 70 percent of its organic search traffic to the same feature. Google invited 20 independent web creators to a summit in October 2024 after algorithm updates collapsed their visibility, then left them without answers for months, while sites like TechRaptor absorbed a further 50 percent visibility loss on the final day of the August 2024 core update. None of these operations held an archive Google needed for Gemini training. None had anything to withhold.

Crawl economics reach the smallest budgets first. Bots accounted for 57.4 percent of web traffic to HTML content by June 2026, and Cloudflare's attribution dashboard documented crawl-to-referral ratios approaching 50,000 fetches for a single human visit. PatronView, a site run independently, watched unique daily IP addresses climb from roughly 18,000 to 124,000 in late July 2026, traced to a residential proxy network, and blocked Amazon's crawler after logging 117,000 page reads per day; the same article records Anthropic's crawler returning one referral for every 38,000 pages fetched as of July 2025. Residential proxies that dress automated traffic as human visitors belong to the same category of technique Google alleges SerpApi deploys against SearchGuard, a defense Google's own complaint prices at tens of thousands of person hours and millions of dollars. A one-person site meets the same traffic with a hosting bill.

The exits are narrow. Blocking Google's AI use means leaving search altogether, since one crawler serves both functions, and Google's stated compliance with publisher opt-out obligations begins only in the United Kingdom. Standardized licensing, the route that would let a small site sell what Reddit sells, is the proposal Google's Nick Fox rejected in December 2025. The courtroom has been no kinder at this scale: the Helena World Chronicle plaintiffs, small newspaper publishers, saw their antitrust case dismissed for standing in April 2026, while Penske Media's proceeds before the same judge. The statutory remedy this case may sharpen, $200 to $2,500 per circumvention, presumes a plaintiff able to fund federal litigation and a defendant able to pay a judgment, and Google's own complaint argues SerpApi cannot cover the liability it already faces. What Reddit obtained through contract, and Google defends through SearchGuard, small publishers meet with defaults, closures and a doctrine that so far prices their pages at nothing.

Why this matters for marketers

SerpApi's output feeds the rank trackers, share-of-voice dashboards, competitive intelligence platforms and AI-visibility tools that search marketers use daily. The supply chain behind that data has been tightening for nearly a year. Google eliminated the n=100 results parameter on September 14, 2025, sharply raising collection costs for SEO platforms, and restricted SerpApi's Light Fast API workaround to three organic results in early October 2025. A revived DMCA theory would add statutory copyright exposure, of $200 to $2,500 per query, on top of those technical constraints for any provider whose collection touches licensed content in results. A second dismissal would leave the July finding in place: no new copyright barrier for the tools the industry depends on. SerpApi's response to the amended complaint will set which of those worlds arrives first, and publishers building their own paid-access walls will be reading the docket as closely as the SEO industry.

Timeline

Summary

Who: Google LLC, represented by Wilson Sonsini Goodrich & Rosati, against SerpApi LLC, an Austin, Texas API provider, before Chief Judge Yvonne Gonzalez Rogers in the United States District Court for the Northern District of California, with Reddit, Cloudflare, publishers and search data providers positioned around the outcome.

What: Google filed a 15-page amended complaint reasserting two DMCA anti-circumvention claims against SerpApi, this time grounded in licensing agreements, including Reddit's directive that Google not enable third parties to extract and commercialize licensed content, after the court dismissed the original complaint for failing to allege copyright-owner authorization behind SearchGuard.

When: The amended complaint was filed on August 10, 2026, the final day of the 21-day amendment window set by the July 20, 2026 dismissal order. SerpApi counsel responded on August 11, 2026.

Where: Case No. 4:25-cv-10826-YGR in the Northern District of California, Oakland Division, with a parallel Reddit case against SerpApi and Perplexity proceeding in the Southern District of New York.

Why: The court's construction of Section 1201 ties DMCA protection for search results to licensed content and rights-holder authorization, making Google's licensing contracts the load-bearing structure of the case. The outcome will influence whether automated access to search results and licensed web content consolidates into a paid, contract-governed transaction, a direction pay-per-crawl systems, licensing renewals and mandated syndication remedies are already pushing from other angles. The same arguments circulate through publisher litigation against Google itself, from Penske Media's federal antitrust case to the Munich liability ruling, while the pleading elements the order spelled out map onto positions that content-owning publishers of any size can occupy. A revived case would also open the licensing agreements, their negotiation records and SearchGuard's design documentation to discovery, since Google has pleaded that material as the factual core of its claims. Binding orders in Washington and Brussels compel Google to open search data and syndication outward at the same time, completing a squeeze in which the company litigates to punish unlicensed taking while being ordered to license.