Overview
In its decision last month in Receivership Estate of AudienceScience Inc. v. Google LLC[1], the Federal Circuit affirmed that internet ad-selection claims were ineligible under 35 U.S.C. § 101. While the challenged claims did not invoke the use of artificial intelligence or machine learning, the Federal Circuit’s reasoning adds to a line of decisions rejecting claims directed to taking a known business practice and running it on a newer, more sophisticated technology.
I. Application of Alice Two-Step Framework
The Federal Circuit rejected claims “directed to the abstract idea of targeting advertisements to internet users to maximize revenue generation” as unpatentable under the two-step Alice framework.[2]
Under Alice step one, the Court found that a claim directed to a method of “selecting an advertising message for inclusion into a requested web page,” was abstract because the claim “as a whole, [was] directed to the abstract idea of targeting advertisements to internet users to maximize revenue generation.”[3] The Court rejected AudienceScience’s argument that this framing was improperly “untethered from the claims,” because “the claims explicitly recite[d] selecting and displaying advertisements based on information collected about a user and advertisement performance.”[4] Specifically, the Court found that the patent’s specification, which described page context and user history as “conventional approaches” to targeted advertising as supporting the Court’s conclusion that the claimed “abstract idea is not untethered from the claims.”[5]
The Court further rejected AudienceScience’s arguments on why the claims survive Alice step one. First, the Court explained that the claims did not solve “a problem specifically arising in the realm of computer networks” as they did under DDR Holdings, which addressed a “problem that [did] not arise in the ‘brick and mortar’ context.”[6] Instead, the Court found that selecting the most effective advertisement is instead a “non-internet centric problem” analogous to print and television advertising.[7] Second, as in Chewy v. IBM, the Court found that the claims recite gathering keywords, matching advertisements, and displaying one “without providing any specificity as to how those steps are achieved.”[8] The Court explained that the claimed performance score, even if it “add[ed] a degree of particularity,” was “not enough to offset the abstract concept embodied by the claims.”[9] Finally, the Court held that SRI, Finjan, and McRO were distinguishable because they “rested on the recitation of specific solutions for solving technological problems,” whereas “nothing in the claims or specification [in the asserted patents] indicate[d] any specific technological improvements or solutions to an internet-centric problem in advertising.”[10]
At Alice step two, the Federal Circuit affirmed that the asserted claims “lack[ed] an inventive concept,” and rejected both of AudienceScience’s arguments to the contrary.[11] First, AudienceScience argued that “[t]he use of both page context and user history is an inventive concept because it allow[ed] advertisement recommendations to account for both the subject matter that [was] currently on the user’s mind and subject matter that the user ha[d] previously demonstrated an interest in.”[12] The Court disagreed because the specification itself described each targeted advertising method as “conventional.”[13] Second, AudienceScience argued that assigning “performance scores” and weights “in accordance with [those] performance score[s]” solved an internet-specific problem and was adaptive because it used “probabilistic selection.”[14] In rejecting this argument, the Court distinguished Weisner v. Google, which AudienceScience relied on, explaining that in Weisner, the claims had “specificity as to the mechanism through which they achieve improved search results.”[15] By contrast, the claims of the asserted patent “simply recite[d] ‘attributing a selection weighting’” reflecting a performance score tied to revenue, and the claims “[did] not explain in any greater specificity how this performance score concept [was] carried out.”[16]
II. Three Points That Carry Over to AI Claims
1. Applying New Technology to an Abstract Idea Does Not Establish Patent Eligibility
AudienceScience argued that using page context, browsing history, and ad performance together was “not possible before the Internet.”[17] The Court rejected this argument explaining that selecting effective advertisements is a “non-internet centric problem,” and publishers have long done it in print and on television.[18] The Court held that applying a “known business process to the particular technological environment of the Internet” did not confer patent eligibility.[19] An AI-based version of a long-standing business practice invites the same objection.
2. Better Results Alone Do Not Supply an Inventive Concept.
AudienceScience said combining page context and user history produced “far more personalized results.”[20] The Court held this rationale “collapses[d] into the abstract idea itself,” and that the abstract idea cannot supply its own inventive concept.[21] An AI claim that relies solely on a model’s ability to produce better results faces the same problem.
3. Claiming a Result Is Not a Substitute for a Specific Technical Implementation
The Court noted the claims at issue did not explain “in any greater specificity how this performance score concept is carried out.”[22] AudienceScience also argued its score was “adaptive” and used “probabilistic selection.”[23] The Court was not persuaded because the claims and specification did not tie that feature to a technical mechanism. Courts are likely to be equally unreceptive to claims that invoke a generic model or algorithm as a black box that delivers the desired outcome.
III. How the Court Distinguished Eligible Claims
The Court contrasted SRI, Finjan, and McRO, explaining that those claims recited specific techniques that solved technological problems or improved computer functionality. The same dividing line will play a role in AI claims. Claims directed to specific training methods or techniques that target technological problems specific to a particular domain or architecture will stand on firmer ground. Claims merely applying generic predictive tools to long-standing business practices or principles are less likely to pass muster.
[1] The Receivership Estate of Audiencescience Inc., Revitalization Partners, LLC, Plaintiffs-Appellants v. Google LLC, YouTube LLC, Defendants-Appellees, No. 2024-1825, 2026 WL 2880090, at *1 (Fed. Cir. Sept. 25, 2026).
[2] Id.
[3] Id. at 4.
[4] Id.
[5] Id.
[6] Id. at 5 (citing DDR Holdings, LLC v. Hotels.com, L.P., 773 F.3d 1245, 1257 (Fed. Cir. 2014)).
[7] Id.
[8] Id. (citing Chewy, Inc. v. International Business Machines Corp. 94 F.4th 1354 (Fed. Cir. 2024)).
[9] Id.
[10] Id. at 6 (citing SRI Int'l, Inc. v. Cisco Sys., Inc., 930 F.3d 1295 (Fed. Cir. 2019); Finjan, Inc. v. Blue Coat Sys., Inc., 879 F.3d 1299 (Fed. Cir. 2018); McRO, Inc. v. Bandai Namco Games Am. Inc., 837 F.3d 1299 (Fed. Cir. 2016)).
[11] Id. at 7.
[12] Id. at 6.
[13] Id.
[14] Id.
[15] Id. (citing Weisner v. Google LLC, 51 F.4th 1073, 1085 (Fed. Cir. 2022)).
[16] Id. at 7.
[17] Id. at 5.
[18] Id.
[19] Id.
[20] Id. at 6.
[21] Id.
[22] Id. at 7.
[23] Id. at 6.
