Google LLC v. Parus Holdings — Federal Circuit Vacates IPR Decision, Faults PTAB for Analyzing Prior Art in Isolation and for Secret Claim Construction

Background

Parus Holdings, Inc. owns U.S. Patent No. 6,721,705, titled “Robust Voice Browser System and Voice Activated Device Controller.” The patent describes a telephone-based system for browsing the Internet by voice: a user calls in, the system accepts voice commands, retrieves information from ranked websites, and reads the results back to the user. A key feature is a “polling mechanism” that periodically tests each website’s responsiveness and adjusts its ranking accordingly — moving slow, broken, or changed sites down in priority so that the system preferentially queries reliable sites first.

Google filed a petition for inter partes review (IPR) of the ‘705 patent in December 2021, asserting that claims 1 and 2 were obvious over two combinations of prior art: (1) Kovatch (an IVR system that guides users through a voice-navigated index of Internet destinations) combined with Burrows (a patent on modified collection frequency ranking for search results); and (2) Wise (a voice-operated computer search engine) combined with Burrows. In August 2024, the Patent Trial and Appeal Board (PTAB) issued its final written decision upholding the challenged claims, finding that Google had not shown the claimed polling-and-ranking features would have been obvious.

This decision is related to but separate from an earlier LexSummary post about Parus/Google litigation: in 2023, the Federal Circuit addressed PTAB’s procedural rules on evidence incorporated by reference in the related ‘941 and ‘402 patent IPRs. In those earlier IPRs, the Board had actually found similar Parus claims unpatentable for obviousness over Wise and Burrows — which set the stage for today’s inconsistency argument.

The Court’s Holding

The Federal Circuit vacated the Board’s decision and remanded for further proceedings on three independent grounds.

1. The Board analyzed prior art references in isolation, not in combination. Google’s petition asserted that the combination of Kovatch (or Wise) with Burrows would have rendered the claims obvious. But the Board’s analysis focused on whether Burrows, standing alone, disclosed the disputed claim limitations — specifically the polling mechanism’s features for decreasing a website’s rank when it is nonresponsive [limitation 1H], returns an unexpected response [1I], or responds too slowly [1J]. The Board asked “does Burrows teach X?” rather than “would a skilled artisan have been motivated to modify Kovatch or Wise based on Burrows’s teachings to achieve X?” The Federal Circuit held this was legal error: prior-art references in an obviousness combination “must be read together, not in isolation” when the petitioner’s theory relies on how the references combine. On remand, the Board must evaluate the specific Kovatch-Burrows and Wise-Burrows combinations as Google’s petition presented them.

2. Undisclosed claim construction violated the APA. For limitation [1I] — requiring the polling mechanism to decrease a site’s ranking “if an unexpected response is received” — the Board adopted a construction of “unexpected response” that excluded routine changes to webpage content. But the parties had not contested the meaning of this term as a claim-construction matter, so Google had no notice that the Board intended to construe it. The court found that the Board’s reasoning — citing the patent specification’s preferred embodiment and its own assessment of what webpage changes “could not” be unexpected — amounted to implicit claim construction. Under the APA, a PTAB petitioner is entitled to notice of and an opportunity to respond to any new claim construction. Because Google was denied that opportunity, the Board’s analysis of limitation [1I] must be redone on remand with proper notice.

3. The Board failed to address an inconsistency with earlier rulings. In the parallel IPRs of two related Parus patents (the ‘941 and ‘402 patents), the Board had found that the Wise-Burrows combination rendered similar polling-and-ranking claims unpatentable. Google raised those earlier decisions in its reply brief and at oral argument, arguing that the Board could not logically uphold the ‘705 claims over Wise-Burrows while having found essentially identical features obvious in the related IPRs. The Board never addressed this argument. The court held that the APA requires an agency to explain when it reaches apparently discrepant results in similar cases — and the Board’s complete silence made meaningful appellate review impossible. If the Board again upholds the ‘705 claims on remand, it must explain why this case is relevantly different from the ‘941 and ‘402 IPRs.

Key Takeaways

• Obviousness combinations must be analyzed as a whole: the question is whether a skilled artisan would be motivated to modify one reference based on another’s teachings, not whether each reference independently discloses every element.
• The APA’s notice-and-comment requirements apply to PTAB claim construction even when neither party explicitly framed a dispute as a claim-construction fight — if the Board’s analysis resolves the scope of a claim term, that is claim construction and requires notice.
• The Board cannot reach inconsistent conclusions across related IPRs without explaining the difference — agency adjudicators have an APA obligation to provide reasoned explanations for discrepant outcomes in similar cases.
• Google wins a remand but not a reversal — the Board may still uphold the ‘705 patent claims if it corrects its analytical approach on remand.

Why It Matters

This case is part of a broader line of Federal Circuit decisions reinforcing procedural rigor at the PTAB. The court has now repeatedly found that the Board violates the APA when it construes claims without warning, ignores inconsistencies with its own prior decisions, or focuses too narrowly on individual prior-art references rather than the petitioner’s theory of how they combine. For IPR petitioners, the decision underscores the importance of clearly articulating the theory of how combined references teach the challenged limitations — and of preserving the record on any potential inconsistency between the Board’s findings in related proceedings.

For patent owners defending in IPR, the decision is a reminder that the Board’s procedural missteps can be a double-edged sword: here, a Parus “win” at the PTAB was vacated not because Google was right on the merits, but because the Board failed to follow proper procedure. The patent now faces a do-over at the PTAB under a corrected framework.