Google’s Knowledge Panel Fix Risks Its Own Search Defenses

Google confirmed it will amend its DMCA complaint against SerpApi after a federal judge dismissed the original claims on July 20, 2026. The new focus lands on licensed media inside Knowledge Panels, a narrow path that keeps the anti-scraping fight alive while exposing Google to fresh legal risks of its own.

U.S. District Judge Yvonne Gonzalez Rogers of the Northern District of California ruled that plain search results carry no copyright and that SearchGuard, Google’s JavaScript challenge system, cannot invoke the DMCA where no protected work exists. Google now has 21 days to refile.

Court Cuts the Broad Claim Clean

Google filed suit to stop SerpApi bots on December 19, 2025. The complaint alleged SerpApi circumvented SearchGuard, launched in January 2025, then resold structured Google results through an unauthorized API. SerpApi’s clients include Nvidia, Uber and Adobe.

The judge accepted that Google may have standing even without owning the underlying content. Yet she held the DMCA requires a technological measure that effectively controls access to a work protected by copyright. Plain SERP listings of public links and snippets do not qualify.

1. December 19, 2025, Google files DMCA complaint against SerpApi.
2. January 2025, SearchGuard JavaScript challenge goes live and briefly blocks SerpApi.
3. February 20, 2026, SerpApi moves to dismiss.
4. July 20, 2026, Court grants dismissal with leave to amend on copyrighted subsets only.

The order granting the motion to dismiss leaves Google one narrow door: content it licenses and protects with the authority of the copyright owner.

That sequence matters because the court separated two questions the original complaint had fused. Standing to sue survived. Protection under the DMCA for ordinary result pages did not. The leave to amend is therefore not a second chance at the same theory. It is permission to plead a smaller set of works that actually meet the statute’s copyright threshold.

SearchGuard’s design compounds the problem. The challenge system blocks automated access across the board. When the blocked material includes no copyrighted work belonging to Google or its licensors, the circumvention claim collapses. The judge’s order simply applied that statutory filter to the facts Google itself had alleged.

Knowledge Panels Become the Needle

Google’s amended theory centers on Knowledge Panels. These summary boxes sit at the top of many results and pull facts and images from the Knowledge Graph. Some images and media arrive under explicit licenses. Google argues rights holders authorized the company to deploy anti-scraping tools around that licensed material.

A Google spokesperson, José Castañeda, told Ars Technica the company is “pleased to see that the Court rejected nearly all of SerpApi’s legal arguments” on standing and looks forward to the amended filing. The strategy tracks the original complaint’s example of a panel containing one copyrighted image among several open ones.

Official documentation explains how Knowledge Panels display licensed media alongside public data. That mix is exactly why the court rejected a blanket claim: SearchGuard blocks everything, licensed or not.

The amended complaint must therefore isolate the licensed slice with enough precision to satisfy the DMCA while still describing a technological measure that controls access to it. A single copyrighted image inside a panel of otherwise open material becomes the test case the original filing already sketched. Whether that thin example can carry a full claim is the open question the 21-day window now forces Google to answer.

• Licensed images and media inside panels may support a DMCA theory
• Public facts and open snippets inside the same panels do not
• SearchGuard’s uniform block does not itself create copyright protection
• Rights-holder authorization must match the measure actually deployed

The Risk Google Just Bought

Meredith Rose, senior policy counsel at Public Knowledge, told Ars the amendment path is treacherous. Google must argue that certain panel elements are reproductions of copyrighted works it licenses. Anything else risks admitting the company algorithmically reproduces unlicensed content across search features.

They have to make an argument somehow that there are parts of that that are reproductions of copyrighted material, but the only parts that are reproductions of copyrighted material are the ones they’ve explicitly licensed. Otherwise, they’re admitting that they have been reproducing stuff without licensing it, and that gets them into another fair use fight that they probably don’t want to have.

Rose made the warning in the same Ars Technica interview that carried Castañeda’s response. The Catch-22 is real: success on the narrow claim can weaken Google’s long-standing fair-use posture for AI Overviews and other generated summaries.

On X, observers immediately flagged the irony. One widely shared take noted that refiling would require Google to concede site owners hold enforceable copyrights that could later support suits against Google itself for scraping and summarizing their pages.

The tension is structural. Google needs the licensed-content argument to keep a DMCA claim alive. The same argument, once on the record, supplies a template for rights holders who object to how their material appears in other Google products. Castañeda’s focus on standing wins does not dissolve that downstream exposure.

Clients and Reddit Sit in the Blast Radius

SerpApi insists the fight is larger than one vendor. CEO Julien Khaleghy called the dismissal a win not just for SerpApi but open internet. The company said Google and Reddit “do not own the Internet” and should not act as toll collectors for public information.

Reddit’s own suit, filed earlier, faces similar standing problems. Rose noted Reddit is neither copyright owner nor exclusive licensee of user posts that appear in Google results, nor the party that built SearchGuard. A hearing on Reddit’s motion occurred around the same window; the Google order does not help its position.

• AI training pipelines that still rely on fresh ranked results
• SEO and research tools that parse public SERPs at scale
• Publishers weighing whether to lock more content behind walls
• Archivists and journalists who depend on automated crawls

SerpApi told Ars its business kept growing through the litigation, yet customers have lived with legal uncertainty for months. The company framed both Google and Reddit as attempts to re-enclose public data.

Named customers such as Nvidia, Uber and Adobe illustrate the practical stakes. Their pipelines depend on structured results that the court has now held fall outside the original DMCA theory. Continuity of service is intact for the moment. The legal cloud over how that data is obtained has not fully lifted.

Scrapers Who Built Empires Now Litigate

Techdirt and others noted the historical symmetry. Google’s index itself was built by crawling the open web. The company now deploys SearchGuard and DMCA 1201 claims to stop others from doing a narrower version of the same act on its own public pages.

Earlier precedents such as hiQ Labs v. LinkedIn established that scraping publicly available data often falls outside the Computer Fraud and Abuse Act. The DMCA route Google chose is different and, so far, less successful for plain results. The court left only the licensed-content slice open.

That slice may prove too small to justify the cost of continued litigation, or it may become a nuisance pressure tool. Either way, the posture has shifted from “protect hundreds of thousands of publishers” to “protect the licensed images we put in some panels.”

The contrast with hiQ is instructive on method rather than outcome. That line of cases tested access restrictions under a different statute. Google’s choice of the DMCA put the existence of a copyrighted work at the center of the analysis. Once the court found no such work in plain SERPs, the circumvention claim had nowhere to land.

Data Access After the Ruling

The practical effect lands on how AI systems and developers obtain structured search data. If Google succeeds even on the narrow claim, SerpApi and similar services may simply filter out Knowledge Panel payloads. The rest of the public SERP would remain fair game under the current order.

Key numbers from the record:

• 21 days, window Google received to file its amended complaint
• Billions, automated queries Google says SerpApi generates, creating deadweight cost
• Two DMCA claims, both dismissed as originally pled; only copyrighted subsets may return
• One licensed image example, the complaint’s own illustration of mixed panel content

Rose described a broader “re-enclosure” of the web since 2023 as publishers and platforms reach for any available tool against AI-scale scraping. Research, journalism and public-interest crawling feel the same chill. SerpApi positioned itself as the party willing to litigate that principle even at high cost.

Filtering Knowledge Panels would be a technical workaround, not a legal surrender. It would accept the court’s holding on plain results while removing the only content category Google has been invited to replead. Whether that compromise satisfies customers who want full page structure is a business question the litigation has now made concrete.

The Amendment Forces a Narrower Proof

Google’s next filing must do more than restate the original grievance. It has to map specific licensed works to the technological measure that guards them and show that SerpApi’s access defeated that measure as to those works alone.

The complaint’s own illustration of one copyrighted image among open panel elements sets the bar. Authorization from the rights holder must be real, documented, and tied to anti-scraping enforcement. Without that chain, the narrow door the court left open closes again.

Success on those points would still leave the bulk of public results outside DMCA reach under this order. Failure would end the federal copyright theory against SerpApi for now. The 21-day clock makes that binary choice immediate.

How Open Web Arguments Meet Platform Controls

Khaleghy’s framing that Google and Reddit “do not own the Internet” collides with platform efforts to treat their result surfaces as controlled environments. The court did not adopt either slogan. It applied the DMCA’s text to the works Google identified and found the match missing.

That outcome leaves both sides with partial tools. SerpApi keeps the holding that plain results are not protected works under this theory. Google keeps standing and a chance to replead around licensed media. Rights holders gain a reminder that licenses can become enforcement hooks, and also that those same hooks can point back at the platforms that display their work.

The re-enclosure pattern Rose described since 2023 supplies the climate, not the holding. Publishers reach for contracts, walls, and statutes. Scrapers litigate for continued access to public pages. This case shows how quickly a broad statutory claim can shrink once a court demands an actual copyrighted work.

Google’s amendment will clarify how tightly it tries to thread the needle. The filing will also show whether rights-holder authorizations actually exist in the form the statute requires. Until then the broad claim is dead, the narrow claim is untested, and the open-web argument remains the one SerpApi is prepared to keep making.

Platforms face a growing stack of liability fights. One recent example is another high-stakes platform liability case now heading toward trial. The SerpApi matter sits in the same legal climate: large platforms testing novel theories to control how their surfaces are used by third parties.

The amended complaint will tell the next chapter. The irony is already on the page.