Patent opinions from the Eastern District of Texas and the U.S. Court of Appeals for the Federal Circuit

Weekly Digest
Fed. Cir.

Google LLC v. Parus Holdings, Inc., No. 24-2381 (Fed. Cir. July 13, 2026)

Vacated & Remanded
Court
U.S. Court of Appeals for the Federal Circuit
Case No.
No. 24-2381
Decided
July 13, 2026
Judge
Taranto, J.
Document
Nonprecedential Opinion
Length
15 pages

NOTE: This disposition is nonprecedential. United States Court of Appeals

for the Federal Circuit

GOOGLE LLC,

Appellant

v.

PARUS HOLDINGS, INC.,

Appellee

2024-2381

Appeal from the United States Patent and Trademark Office, Patent Trial and Appeal Board in No. IPR2022-00279.

Decided: July 13, 2026

BENJAMIN HABER, O'Melveny & Myers LLP, Los Angeles, CA, argued for appellant. Also represented by WILLIAM FINK, Washington, DC; CAITLIN P. HOGAN, New York, NY. CHARLES E. FOWLER, JR., McKool Smith, P.C., Austin, TX, argued for appellee. Also represented by JOHN BRUCE CAMPBELL; CHRISTIAN DORMAN, Washington, DC.

2Before DYK and TARANTO, Circuit Judges, and MOORE,

District Judge.1

Parus Holdings, Inc., owns U.S. Patent No. 6,721,705, which describes and claims a voice-operated system for browsing the Internet. As relevant here, Parus also owns two other patents descended from the ’705 patent. Google LLC successfully petitioned the Patent and Trademark Office (PTO) for inter partes reviews (IPRs) of the ’705 patent and related Parus patents under 35 U.S.C. §§ 311–19. In its petitions, Google challenged various claims on obviousness grounds, asserting similar prior-art combinations and rationales for obviousness. In the two non-’705 IPRs of relevance, the PTO’s Patent Trial and Appeal Board held all challenged claims unpatentable for obviousness, and we summarily affirmed those Board decisions in 2025. In contrast, in the IPR of the ’705 patent, which is now before us, the Board determined that Google had failed to show the unpatentability of the claims at issue (claims 1 and 2). Google LLC v. Parus Holdings, Inc., No. IPR2022-00279, 2024 WL 3625783, at *1 (P.T.A.B. Aug. 1, 2024) (’705 Decision). Google appeals. We vacate the Board’s decision and remand for further proceedings.

I

A

The ’705 patent, which issued from a 2001 nonprovisional application (following two provisionals filed in 2000), is titled “Robust Voice Browser System and Voice Activated Device Controller.” It has a number of descendants, including U.S. Patent Nos. 7,881,941 and 8,185,402, also owned3 by Parus, with which it shares a specification. The ’705 patent describes a need “for a system that allows users to easily access and browse the Internet” using “any type of telephone.” ’705 patent, col. 2, lines 19–23. The patent purports to meet that need with a telephonic “voice browsing system” that allows users to control a web browser with voice commands. Id., col. 2, lines 47–63.

The described system maintains categories of information that might be sought by a user and, within each category, a ranked list of websites. See id., col. 5, lines 7– 26. When a user requests a particular type of information (e.g., an extended weather forecast), the system accesses the pertinent websites in the order that they are ranked to attempt to find responsive information. See id.; id., col. 16, lines 23–35. The system updates its ranked lists by using a “polling mechanism” that “continually polls or ‘pings’ each of the sites.” Id., col. 16, lines 48–52. The website rankings are adjusted based on whether a site provides a “useful and error-free” response to the poll and on its “speed . . . in providing responses.” Id., col. 17, lines 5–20. This ranking method, according to the specification, ensures “complete, timely, and relevant responses.” Id., col. 17, lines 11–13.

The ’705 patent has four claims, all of which are independent, but only claims 1 and 2 are involved in this appeal. Claims 1 and 2 are similar and the parties treat claim 1 as representative. Claim 1 reads as follows (with clause labels as added in the Board proceeding):

1. An Internet voice browsing system for gathering information from web sites on the Internet, comprising: [1A] a CPU-based media server, said media server including at least a speech recognition engine, a speech synthesis engine, an interactive voice response application, a call processing system, and telephony hardware, said media server configured
4to receive a speech command from a user and to convert said speech command into a digital data message; [1B] a database containing a list of web sites stored on magnetic media, [1C] a rank number assigned to each one of said web sites and stored in said database; [1D] a CPU-based web browsing server, said web browsing server including at least a content extraction agent, a content fetcher, a polling and ranking agent, and a content descriptor file, [1E] said web browsing server configured to receive said digital data message from said media server and configured to access one of said web sites having the highest said rank number and to retrieve information from said one of said web sites; [1F] said media server configured to generate an audio message representing said information and to transmit said audio message to said user; and [1G] a polling mechanism configured to periodically send a polling digital data message to each one of said web sites and to receive a response, each of said web sites thereby becoming a polled web site, [1H] said polling mechanism configured to decrease said rank number of said polled web site if no response is received from said polled web site, [1I] said polling mechanism also configured to decrease said rank number of said polled web site if an unexpected response is received from said polled web site, and [1J] said polling mechanism also configured to decreases said rank number of said polled web site if a response time of said polled web site is longer than a second response time of a second polled web site.

5Id., col. 19, line 47, through col. 20, line 17 (emphasis added).

B

Google filed a petition for an IPR of all claims (1 through 4) of the ’705 patent in December 2021. J.A. 125, 207. Google eventually dropped claims 3 and 4 from the IPR. As to claims 1 and 2, Google argued unpatentability for obviousness, as explained in greater detail infra, over two combinations of references: Kovatch in view of Burrows and Wise in view of Burrows. J.A. 125. In January and February 2022, Google also sought IPRs of all claims of the ’941 and ’402 patents; in those petitions, Google alleged the obviousness of certain claims over a combination including Kovatch and Burrows as well as a combination including Wise and Burrows. See J.A. 2817, 3445.

Kovatch, titled “Interactive Voice Response System,” is a Patent Cooperation Treaty application, No. WO 01/50453. J.A. 364. It describes “an interactive voice response system” that permits navigation of a computer using only voice commands by guiding a user through an “index of destinations.” J.A. 368, lines 23–27. Wise is U.S. Patent No. 5,884,262, which is titled “Computer Network Audio Access and Conversion System” and discloses a voice-operated search engine that makes use of an index of audio-compatible documents. See J.A. 477, col. 9, line 56, through col. 10, line 10. Burrows is U.S. Patent No. 5,765,149, which is titled “Modified Collection Frequency Ranking Method” and describes a computerized method for ranking records (e.g., webpages) in response to a user’s search using a preconstructed index of search terms and locations. See J.A. 452, col. 1, line 53, through col. 2, line 21.

In July 2022, the Board denied institution of the ’705 IPR. J.A. 1387. Google sought rehearing that month. J.A. 1422–35. While Google’s rehearing request was pending, the Board instituted the ’941 and ’402 IPRs, and in August6 2023, the Board held all challenged claims of the ’941 and ’402 patents unpatentable on the Wise grounds without reaching the grounds involving Kovatch. J.A. 2308–45, 2346–83.2 Then, in September 2023, the Board granted Google’s rehearing request and instituted the ’705 IPR. J.A. 1492.

In August 2024, the Board rendered a final written decision upholding claims 1 and 2. ’705 Decision, at *1. It determined that Google had not shown that claim limitations [1H], [1I], and [1J] (or the corresponding claim 2 limitations) would have been obvious as part of either the Kovatch-Burrows or Wise-Burrows grounds and, further, that Google had not shown that claim limitation [1C] (or the corresponding claim 2 limitation) would have been obvious as part of the Wise-Burrows ground. Id. at *7–17.

Google timely appealed. We have jurisdiction under 28 U.S.C. § 1295(a)(4) and 35 U.S.C. §§ 141(c), 319.

II

We review the Board’s decisions under the standards of the Administrative Procedure Act (APA) and must “set aside conclusions or findings that are ‘arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law,’ or ‘unsupported by substantial evidence.’” Alacritech, Inc. v. Intel Corp., 966 F.3d 1367, 1370 (Fed. Cir. 2020) (quoting 5 U.S.C. § 706).

Google argues that the Board made three errors. First, Google contends, the Board legally erred by considering the prior-art references in isolation from each other in finding failure by Google to show obviousness as to claim limitations [1H]–[1J] for both grounds and limitation [1C] for the7 Wise-Burrows ground. Second, Google argues, with respect to both grounds, that the Board denied it the APA-required process by adopting a claim construction of the term “unexpected response” in its decision without having provided adequate notice of or opportunity to address the construction. Finally, for the Wise-Burrows ground, Google contends that the Board failed to address Google’s argument that upholding the ’705 claims would be inconsistent with the Board’s unpatentability determinations in the ’941 and ’402 IPRs. We are persuaded that these errors require vacatur and remand.

A

1

As to both the Kovatch-Burrows and Wise-Burrows combinations, Google contends that the Board focused too narrowly on Burrows in isolation when finding that Google failed to prove the challenged claims obvious. We agree, and we conclude that a remand is necessary for the Board to find facts with a proper focus on the combinations as presented in the petition.

Google’s petition asserted that both Kovatch in view of Burrows and Wise in view of Burrows would have rendered claims 1 and 2 obvious. J.A. 135–37, 157–58. We start with the Kovatch-Burrows combination.

Kovatch, like the ’705 patent, relates to a voice-operated system for accessing information on the Internet. See J.A. 368–69. It discloses organizing possible locations of user-sought information (websites) into an index of destinations through which it guides a user. J.A. 368, 404. As now relevant, Google asserted that Kovatch discloses ranking websites, as required by limitation [1C], because, Google said, Kovatch teaches indexing websites in “a hierarchical order based on users’ preferences.” J.A. 144.

Burrows discusses ranking webpages in response to a user’s search based on how frequently search terms appear8 on each page, with less common terms given greater weight. J.A. 452, col. 1, line 53, through col. 2, line 13. Burrows discloses (1) marking inaccessible webpages as deleted and preventing such deleted pages from appearing in search results, J.A. 459, col. 15, lines 11–24, (2) detecting modifications to webpages, J.A. 454, col. 5, lines 42–49, and (3) optimizing how user queries are parsed to reduce the time it takes to complete a search, J.A. 466, col. 30, lines 16–53. Google contended that a relevant artisan would have been motivated “to include Burrows’s teachings . . . to perform the rank assignment” in Kovatch’s index of ranked websites. J.A. 149.

About limitation [1H], Google alleged the obviousness of using “Burrows’s teaching of determining whether an indexed page is inactive . . . to keep Kovatch’s index updated, thereby decreasing the ranking of an inactive page.” J.A. 152. As to [1I], Google contended that a relevant artisan would have been motivated to apply Burrows’s teaching of detecting modified webpages to lower the rank of a website “when the page is a ‘modified page’ to further limit the pages that may not be of as much interest,” and that “as applied to Kovatch, this teaching of Burrows would result in modified entries moving to lower ranked positions in Kovatch’s hierarchy.” J.A. 154. For [1J], Google asserted that Burrows’s teaching of optimizing searches would have motivated a relevant artisan “to use response times as a factor to update the index.” J.A. 155. Google’s petition thus advanced a theory that Burrows teaches three concepts—detecting deleted webpages, detecting modified webpages, and optimizing search response times—that a relevant artisan would have been motivated to use, for various asserted reasons, to modify the website ranking allegedly taught by Kovatch.

The Board, in rejecting Google’s arguments, focused exclusively on the disclosure of Burrows. As to [1H] and its requirement of decreasing the rank of a nonresponsive9 page, the Board rejected Google’s arguments because “Burrows’ system overall has no sense of ranking prior to a query” and—in Burrows’s search algorithm—a deleted “webpage’s rank is not decreased.” ’705 Decision, at *9 (emphasis added). For [1I] (in addition to offering a claim-construction rationale discussed infra, Section II.A.2), the Board analyzed the operation of Burrows to conclude that decreasing a modified page’s ranking would not have been obvious, writing, “Burrows’ ranking and, consequently, the ranking in the combined system, would be based on the content of the webpage, not its response as a changed webpage.” Id. at *11 (emphasis added). And for [1J], the Board concluded that “[Google] point[s] to no evidence that query cost has any impact on the ranking Burrows’ query module [ ] assigns to webpages.” Id. at *13 (emphasis added).

As to the Wise-Burrows combination, the parties’ arguments and the Board’s analysis were much the same as for the Kovatch-Burrows combination, except as to limitation [1C]. See id. at *14–17. Wise relates to a telephone-operated computer system, and teaches indexing audio-compatible files, which “index may be stored as one or more documents preferably in a hierarchical order.” J.A. 477, col. 10, lines 7–9; see J.A. 468, abstract. Google identified this disclosure of Wise as itself teaching [1C]’s numbered ranking of stored websites, but in the alternative, Google argued that (a) Burrows teaches numbered ranking of search results and (b) that teaching would have motivated a relevant artisan to modify Wise to include rank numbers “in Wise’s hierarchical order.” J.A. 164–65. The Board rejected that argument “because Burrows’ ranking operation is performed on results after execution of a query, [so the] ranking cannot be applied to Wise’s index.” ’705 Decision, at *15 (emphasis added).

On appeal, Google argues that the Board’s reasoning failed to engage with the petition’s theory of the role of Burrows in each of the combinations, instead rejecting Google’s10 challenge after analyzing Burrows alone and determining that it does not disclose the limitations just discussed. Google’s argument has merit.

References in an obviousness combination “must be read together, not in isolation,” when, as here, “the particular combination argued by the [p]etitioner” calls for such a reading. Palo Alto Networks, Inc. v. Centripetal Networks, LLC, 122 F.4th 1378, 1386 (Fed. Cir. 2024). As exemplified by the emphasized portions of the Board’s decision just recounted, the Board improperly concentrated on the disclosure of Burrows alone, and whether, for limitations [1H]–[1J], Burrows itself discloses adjusting webpage rankings based on inactivity, modification, or response time, rather than whether a relevant artisan would have been motivated to modify Kovatch or Wise’s rankings to meet the limitations given Burrows’s disclosures relating to inactivity, modification, and response time. Similarly, for Wise-Burrows and limitation [1C], the Board incorrectly limited its analysis to whether Burrows itself teaches a ranking of webpages prior to a search, rather than considering how a relevant artisan would have been motivated to modify Wise based on Burrows’s disclosure of numbered ranking. We cannot say that the deficiencies in the Board’s analysis were harmless, so we remand for the Board to consider the particular combinations presented by Google’s petition.

2

Google further argues, regarding limitation [1I] and both of the petition’s obviousness grounds, that the Board construed the term “unexpected response” but failed to afford Google the APA-required notice of and opportunity to address the construction the Board eventually adopted. On this point too, we agree with Google.

Limitation [1I] requires that the claimed polling mechanism be configured to decrease a website’s rank “if an unexpected response is received” from the website. ’70511 patent, col. 20, lines 10–13. Google, in its petition, alleged that this limitation would have been obvious over Kovatch-Burrows and Wise-Burrows in part because Burrows teaches detecting modifications to webpages and a relevant artisan “would have understood that Burrows’s modified pages are pages that return an ‘unexpected response.’” J.A. 153; see J.A. 174. Parus responded that “the Burrows index has no expectations regarding the content of a particular web page.” J.A. 1634.

The Board rejected Google’s argument. In addition to agreeing with Parus about the disclosure of Burrows (as discussed supra, Section II.A.1), the Board reasoned that Google had not shown “that a change to a webpage would be an ‘unexpected response.’” ’705 Decision, at *10 (analyzing Kovatch-Burrows theory); see id. at *16 (same for Wise-Burrows). It cited the ’705 patent’s disclosure that “web sites change frequently” for that understanding, as well as the “preferred embodiment” of the ’705 patent, which “describes an unexpected response as . . . incomplete or not in the format the server expects.” Id. at *10 (citing ’705 patent, col. 2, lines 24–28; col. 16, lines 54–59). The Board also relied on Google’s expert evidence and a technical publication as support for the idea that changes to webpages were common, so would not be understood to be unexpected. ’705 Decision, at *10 (citing J.A. 249–51; J.A. 901).

Google now contends that the Board went beyond finding facts to construe the claim term “unexpected response” and that Google was not afforded the requisite notice of and opportunity to address that claim construction. In a case like this one where the Board did not expressly interpret the claim language, “[t]o determine whether . . . the Board[ ] has construed a claim, it is helpful to look to the outcome of [its] analysis.” Google LLC v. EcoFactor, Inc., 92 F.4th 1049, 1055 (Fed. Cir. 2024). Where the analysis “establishes the scope . . . and meaning of the patented subject matter,” the Board “has most[ ] likely construed the claim.” Id. (emphasis removed). The Board’s reliance on12 the types of evidence typically used to interpret claims can support a determination that it engaged in claim construction. See id. at 1056.

The Board, in this case, construed the “unexpected response” term. It looked to the ’705 patent’s “preferred embodiment” to determine that “the ’705 patent describes an unexpected response” as a response that “is incomplete or not in the format the server expects,” which “is very different from the routine occurrence of a webpage’s content changing over time.” ’705 Decision, at *10. And upon considering the patent’s specification and extrinsic evidence about the dynamic nature of webpages, the Board seemed to reason that “a change to a webpage,” categorically, could not be the claimed “unexpected response.” Id. That analysis, in reliance on conventional sources of a patent claim’s meaning, restricted the scope of the claim term to exclude changes to webpages, so was claim construction.

We will not address the merits of the Board’s construction ourselves. The Board did not give Google adequate notice of or opportunity to address that claim construction before adopting it. We have repeatedly held that, under the APA, “a petitioner is entitled to respond to a new claim construction.” Axonics, Inc. v. Medtronic, Inc., 75 F.4th 1374, 1382 (Fed. Cir. 2023). That requirement may be satisfied if the Board’s ultimate construction is sufficiently within the range of possibilities raised by a construction dispute between the parties, including if the petitioner has in substance contested the Board’s ultimate interpretation of a claim term, or if there is otherwise an apparent reason that the petitioner should have anticipated the Board’s construction. See id. at 1381 n.7; EcoFactor, 92 F.4th at 1057; Qualcomm Inc. v. Intel Corp., 6 F.4th 1256, 1262–63 (Fed. Cir. 2021). Here, though, the parties did not contest the meaning of “unexpected response” as a matter of claim construction, ’705 Decision, at *4; nor did their arguments reveal different interpretations of the term, see J.A. 153, 1634. Under these circumstances, the Board was required13 to provide Google an opportunity to rebut and to make obviousness arguments under its new interpretation of “unexpected response.”

B

A final reason for vacatur pertains only to the Wise-Burrows ground. Google maintains that it alerted the Board to the earlier ’941 and ’402 IPR decisions and argued that the Board should hold the ’705 claims here unpatentable lest it render a decision inconsistent with its factfinding in those IPR decisions, but that the Board failed to respond to that argument. We are persuaded that the Board insufficiently addressed Google’s assertions based on the earlier IPR decisions.

The Board has an obligation under the APA to render decisions that are not arbitrary and capricious. 5 U.S.C. § 706. Hand-in-hand with that obligation is a requirement that the Board provide a reasoned explanation for a decision, one sufficient to enable our review of the Board’s rationale and its propriety. See Provisur Technologies, Inc. v. Weber, Inc., 50 F.4th 117, 123 (Fed. Cir. 2022). Consistent with those principles, we have recognized that the APA may be violated when an agency adjudicator treats similar cases differently, at least when it does so without explaining why it has reached apparently discrepant results. See Vicor Corp. v. SynQor, Inc., 869 F.3d 1309, 1321– 22 (Fed. Cir. 2017); Hansen-Sorensen v. Wilkie, 909 F.3d 1379, 1384 (Fed. Cir. 2018) (applying same standard in veterans’ benefits context).

Here, Google, in its petitioner reply, straightforwardly contended that Parus’s arguments for the nonobviousness of the challenged claims “conflict[ed] with fact-findings about Burrows in the [’941 IPR] and [’402 IPR].” J.A. 2096 (emphasis removed). Google reiterated the point at oral argument before the Board, saying that the Board had “already applied Burrows to very similar claim elements in14 the ’941 and ’402 patents” to find that claims of those patents would have been obvious, those earlier “conclusions [were] correct,” and Google’s obviousness theory in the ’705 IPR used Burrows “the same way.” J.A. 2698. In the ’941 IPR decision to which Google referred, the Board found that Wise, Burrows, and another reference would have made obvious a claim to a computer that “periodically polls” websites “to determine the availability of each said web site, the duration of time for each said web site to respond to a request from said computer, and changes to the location of said information to be retrieved from each said web site, said computer creating a ranking of said plurality of web sites based on said periodic polling.” ’941 patent, col. 20, lines 33–40; see J.A. 2337–39. In the ’402 IPR decision, the Board similarly found that a combination including Wise and Burrows rendered unpatentable a claim involving ranking websites “based on [ ] periodic polling” to determine “availability,” “duration of time . . . to respond,” and “changes to the location of [ ] information” for each site. ’402 patent, col. 24, lines 57–64; see J.A. 2379, 2381.

Although the Board was thus aware of Google’s position that the earlier IPR decisions involved relevantly similar claims and Wise-Burrows obviousness theories, it did not distinguish or discuss its earlier findings. See ’705 Decision, at *14–17. Sometimes a reviewing court can properly discern the agency’s rationale even if the agency’s statement is less than perfectly clear, while still fulfilling the judicial duty not to displace an agency’s factfinding or other choices with its own. See Motor Vehicle Manufacturers Association, Inc. v. State Farm Mutual Automobile Insurance Co., 463 U.S. 29, 43 (1983). Here, however, the Board’s silence precludes effective appellate review by making it impossible to determine whether the Board apprehended the inconsistency point and, if it did, why it rejected the argument and whether its rationale was sound. On remand, if the Board reaches the Wise-Burrows ground and again determines that the challenged claims are not15 unpatentable, it should expressly address Google’s inconsistency argument and explain why it thinks this case relevantly different from the earlier IPRs or otherwise justify now reaching a different result.

III

We have considered the parties’ remaining arguments and find them unpersuasive. For the foregoing reasons, we vacate the decision of the Board and remand for further proceedings consistent with this opinion.

The parties shall bear their own costs.

VACATED AND REMANDED

Footnotes

  1. 1 Honorable K. Michael Moore, District Judge, United States District Court for the Southern District of Florida, sitting by designation.
  2. 2 We summarily affirmed those decisions in Parus Holdings, Inc. v. Google LLC, Nos. 23-2296, -2297, 2025 WL 720088 (Fed. Cir. March 6, 2025).

Text extracted from the official PDF and reflowed for reading. It is not a certified transcript — the PDF linked above is the authoritative document.

Analysis is coming soon.

A plain-English account of the issues, facts and holding, written by Patent Case Watch. The opinion itself — its full text, its PDF and everything else on this page — stays free to read for everyone, always.

Cite this opinion

Google LLC v. Parus Holdings, Inc., No. 24-2381 (Fed. Cir. July 13, 2026).

Record ID
CAFC-24-2381-20260713
Permalink
https://patentcasewatch.com/opinions/CAFC-24-2381-20260713

No reporter citation is shown because the official source does not supply one; none is inferred here. Check the opinion and a citator before relying on this form.