NOTE: This disposition is nonprecedential. United States Court of Appeals
for the Federal Circuit
GOOGLE LLC,
Appellant
v.
NOBOTS LLC,
Appellee
2024-1432
Appeal from the United States Patent and Trademark Office, Patent Trial and Appeal Board in No. IPR2022-00940.
Decided: November 20, 2025
NATHAN R. SPEED, Wolf, Greenfield & Sacks, P.C., Boston, MA, argued for appellant. Also represented by ELISABETH H. HUNT.
STEPHEN D. ZINDA, Cabello Hall Zinda PLLC, Houston, TX, argued for appellee. Also represented by JAMES H. HALL.
2Before MOORE, Chief Judge, TARANTO, Circuit Judge, and
TARANTO, Circuit Judge.
Nobots LLC is the owner of U.S. Patent No. 9,595,008, titled “Systems, Methods, Apparatus for Evaluating Status of Computing Device User.” Google LLC successfully petitioned the Patent and Trademark Office (PTO) to institute an inter partes review (IPR), under 35 U.S.C. §§ 311–19, of all twenty claims of the ’008 patent. The PTO’s Patent Trial and Appeal Board (Board), upon conducting that review, rejected Google’s challenges to claims 18 and 19 while holding all other claims to be unpatentable. Google LLC v. Nobots LLC, No. IPR2022-00940, 2023 WL 8269284, at *24 (P.T.A.B. Nov. 29, 2023) (Final Written Decision). Google appeals the upholding of claim 19, whose patentability is the only issue before us.
Google disputes the Board’s claim construction of the claim phrase “acquiring interest data.” Nobots defends the Board’s construction and does not dispute that unpatentability follows if we reject the Board’s construction in favor of Google’s position. We hold that the Board’s claim construction was erroneous, and we therefore reverse the Board’s determination that claim 19 is not unpatentable.
I
A
The ’008 patent describes and claims methods of detecting whether a computing device accessing a remote server (to visit a website the server hosts, e.g.) is being operated by a person or an automated program (robot or “bot”). ’008 patent, col. 2, lines 12–18; id., col. 1, lines 10–3 13. The patent describes the prior-art Completely Automated Public Turing test to tell Computers and Humans Apart (“CAPTCHA”), which identifies a user as a “human” or “bot” based on whether the user’s response to a query is correct or incorrect. Id., col. 1, lines 22–27; id., col. 2, lines 25–29. The patent’s proposed advance involves comparing various attributes of a user to those of an expected human user, and, in most embodiments, computing a likelihood that a device is bot-operated, with access denied if that likelihood exceeds a specified threshold. Id., col. 2, lines 18–25.
In one of several categorizations the patent makes for data used for its process, the patent distinguishes “passive data” from “active data.” See id., Abstract. It identifies passive data as data typically stored onto a user’s device or transmitted to a remote location, such as the internet protocol address or geographic location of the device. See id., col. 3, lines 9–24. And it identifies active data as data not typically so stored or transmitted, such as a user’s keystrokes or time of response to stimuli. See id., col. 2, line 60, through col. 3, line 5; id., col. 3, lines 24–32.
Over three consecutive paragraphs, the patent defines the following six two-word phrases: “model data,” “available data,” “acquired data,” “issued data,” “monitored data,” and “interest data.” Id., col. 2, line 54, through col. 3, line 46. The section starts with “[a]s used herein,” id., col. 2, line 54, and concludes with, “[w]ith the foregoing definitions in mind,” id., col. 3, line 47. The patent places each phrase in quotation marks to confirm that lexicography is afoot. E.g., id., col. 2, line 54. For each of the six phrases, the patent extends the definition to “its equivalents and verb forms,” e.g., id., col. 3, line 25, without specifying what constitutes either an equivalent or a verb form.
The patent defines “model data” (and “its equivalents and verb forms”) as comprising attributes of a hypothetical human user accessing a website—providing comparators for evaluating the scrutinized user. Id., col. 2, lines 20–21,4 54–58. Those attributes include both active data and passive data, which, when used as model data, are “active model data” and “passive model data.” Id., col. 2, line 58, through col. 3, line 17.
The “available data” and “acquired data” phrases refer to the scrutinized-user side of the comparison—information about that user’s device or interaction with the server the user seeks to access. Id., col. 3, lines 18–32. The patent defines “available data” (and “its equivalents and verb forms”) and says that this category is a type of passive data. Id., col. 3, lines 18–24. The patent defines “acquired data” (and “its equivalents and verb forms”), which it says is a type of active data, as follows:
data associated with a computing device’s operation and its interaction with a computing environment, such as the Internet, that is generally not recorded within the computing device and/or by other devices that have been affected by the computing device’s operation, but at least some data of which has/have been recorded and/or transmitted to a remote location, such as a server—this is a type of active data.
Id., col. 3, lines 24–32 (emphasis added).
In the Summary of the Invention, the patent describes comparing “acquired and/or available data” with “model data.” Id., col. 2, lines 18–21. In most embodiments, that comparison yields a probability whether the user is a bot. Id., col. 2, lines 21–25.
In the definitions section, the comparison process is further categorized. The patent defines “issued data” (and “its equivalents and verb forms”) as comprising data generated by the server or another device distinct from the scrutinized user’s device. Id. col. 3, lines 33–37. It then defines “monitored data” (and “its equivalents and verb forms”) as comprising “active or passive data, whether available or acquired, obtained” from the scrutinized user’s5 device or external interactions “after the generation of issued data.” Id., col. 3, lines 37–41. Finally, the patent defines “interest data” (and “its equivalents and verb forms”) as comprising the following:
active or passive data, whether available or acquired, that correlates to any data within model data, whether obtained prior to or after the generation of issued data. Thus, interest data includes time independent available data and acquired data, unless qualified differently.
Id., col. 3, lines 41–46 (emphases added).
Immediately after the three definition-stating paragraphs, the patent identifies two sets of embodiments it contemplates. In the first set,
a comparison between interest data, acquired prior to delivery of issued data to the client device, and model data is performed to ascertain the likely status of the client computing device, i.e., human user or bot[.]
Id., col. 3, lines 49–53 (emphasis added). In the second set, the same comparison is done for “monitored data, by definition acquired after delivery of issued data” to the scrutinized user’s device. Id., col. 3, lines 53–58. “In both series of embodiments,” the patent adds, “acquired and/or available data may be used for comparison with suitable model data.” Id., col. 3, lines 58–60.
Independent claim 19 recites: 19. A method for assessing a confidence level that an operator of a client computing device interacting with a server is a human being rather than an autonomic computer application, the method comprising:
6a) acquiring interest data from the client computing device prior to delivery of issued data by the server to the client computing device; b) comparing the interest data to model data relating to human interaction with a computing device prior to the time in which the interest data is acquired; and c) generating a value that represents a confidence level that a human user rather than an autonomic user operated the client computing device prior to the time in which the interest data is acquired.
Id., col. 6, lines 44–57 (emphases added as relevant to the challenged claim construction).
B
In April 2022, after having been sued for infringement by Nobots, Google petitioned the PTO for institution of an IPR of the ’008 patent, challenging all twenty claims. As relevant here, one ground invoked U.S. Patent Application No. 2008/0114624 (Kitts) to render claim 19 unpatentable for anticipation or obviousness. Kitts, titled “Click-Fraud Protector,” describes a method of using data attributes to estimate a likelihood of whether a user seeking access to a website on a remote server is a bot. J.A. 585; 587, fig. 2; 588, fig. 3; see J.A. 593 ¶ 9; 594 ¶ 19. The attributes may include the user’s internet protocol address and residence, J.A. 596–97 ¶¶ 46–48, which are “passive data” under the ’008 patent.
After the Board instituted the IPR, Nobots sought to overcome Kitts by arguing that claim 19’s reference to “acquiring interest data” requires that at least some of the interest data acquired be “active data,” even though the ’008 patent defines “interest data” as “active or passive data, whether available or acquired data.” J.A. 1220–26. According to Nobots, the definition of the phrase “acquired7 data,” which limits that phrase to active data, carries over to limit the three-word verb-with-object phrase “acquiring interest data” to require that at least some active data be acquired. J.A. 1226–28. Google disagreed, arguing that the verb “acquiring” has its ordinary meaning and the verb’s object, the two-word phrase “interest data,” is expressly defined such that wholly passive data qualifies, so the three-word verb-and-object phrase is properly construed to mean the process of getting any data, including passive data, that otherwise qualifies as “interest data” (like an internet protocol address or residence information, as in Kitts). J.A. 1378–82.
The Board agreed with Nobots about the construction of “acquiring interest data” in its final written decision, issued on November 29, 2023. Final Written Decision, at *6– 8. In an analysis of the Kitts reference not challenged here, see Google Opening Br. at 45; Nobots Response Br. at 12– 13, the Board found that Kitts discloses obtaining exclusively “passive data” under the ’008 patent’s definition. Final Written Decision, at *16, 24. The Board, applying its construction of “acquiring interest data” as requiring some active data, concluded that Google had not shown that Kitts discloses or rendered obvious that element of claim 19 and thus had failed to show claim 19’s unpatentability. Id. at *23–24.
Google timely appealed on January 30, 2024. We have jurisdiction under 35 U.S.C. §§ 141(c), 319 and 28 U.S.C. § 1295(a)(4)(A).
II
On appeal, Google asserts that the Board erroneously construed “acquiring interest data” in its determination that claim 19 is not unpatentable. In this case, intrinsic evidence is determinative of the proper claim construction. Accordingly, we decide the claim-construction issue de novo. Intel Corp. v. Qualcomm Inc., 21 F.4th 801, 808 (Fed.8 Cir. 2021). We hold that the Board erred in its claim construction.
“We generally give words of a claim their ordinary meaning in the context of the claim and the whole patent document.” World Class Technology Corp. v. Ormco Corp., 769 F.3d 1120, 1123 (Fed. Cir. 2014); see Thorner v. Sony Computer Entertainment America LLC, 669 F.3d 1362, 1365 (Fed. Cir. 2012). “When a patentee explicitly defines a claim term in the patent specification, the patentee’s definition controls.” Martek Biosciences Corp. v. Nutrinova, Inc., 579 F.3d 1363, 1380 (Fed. Cir. 2009); see Continental Circuits LLC v. Intel Corp., 915 F.3d 788, 796 (Fed. Cir. 2019). To overcome a clear ordinary meaning, the specification must be “clear” in giving a contrary definition or disclaimer. World Class, 769 F.3d at 1123; Thorner, 669 F.3d at 1365. Application of those principles here leads to a straightforward construction.
The three-word phrase, “acquiring interest data,” consists of two parts: the verb “acquiring” and its object “interest data.” It is proper to divide the phrase in that way. The syntactic structure is one of verb and object. There is evident parallelism with the language of the other steps of the method—“acquiring” (interest data), “comparing” (to model data), and “generating” (a confidence level). And the phrase as a whole has no familiar, ordinary meaning as a unit that is different from the sum of its two parts. Cf. Intel Corp. v. Qualcomm Inc., 21 F.4th 784, 791–92 (Fed. Cir. 2021) (citing FCC v. AT&T, 562 U.S. 397, 406 (2011)).
The verb “acquiring” has a plain and ordinary meaning of obtaining, getting, gaining possession, or the like (all relevantly synonymous). It is not limiting as to what must be obtained. That question is addressed in the phrase’s second part, “interest data,” which is a two-word noun phrase and which has no ordinary meaning. It is a phrase that is expressly defined in the specification. See ’008 patent, col. 3, lines 41–46. Under that definition, which governs, “interest data” undisputedly can be “passive data.” Id., col. 3,9 line 42 (“active or passive data”). The overall three-word phrase in dispute simply means gaining possession of whatever is within the category of “interest data,” which can be passive data alone.
There is no sound basis for departing from that straightforward analysis. Critically, this three-word-phrase is not defined in the specification. As relevant here, “interest data” is defined, and “acquired data” is defined, but “acquiring interest data” is not. At a minimum, the high standard for redefinition is not met for this three-word phrase—at all or, more particularly, to replace the meaning it clearly has as a phrase joining an ordinary verb with a defined term.
The specification strongly undermines a contrary conclusion. It uses “acquired” (not followed by “data”) in its ordinary gained-possession sense. In particular, it uses the free-standing word when it describes two sets of embodiments, respectively involving “interest data” and “monitored data,” ’008 patent, col. 3, lines 48–58, having just defined those phrases broadly, each one covering “active or passive data,” id., col. 3, lines 37–46. It is implausible that the patentee, by using the ordinary verb “acquire,” was suddenly cutting back on the breadth just asserted definitionally. Indeed, that usage naturally means the same thing as “obtained” used in the definitions of “monitored data” and “interest data.” Id., col. 3, lines 39, 44.
The patent also affirmatively declares that in “both embodiments” (for monitored data and interest data), “acquired and/or available data”—the first active, the second passive—“may be used for comparison with suitable model data.” Id., col. 3, line 58–60 (emphasis added). The Summary of the Invention makes the same point using the same “and/or” language. Id., col. 2, lines 18–25. That language, in its ordinary meaning, says that either acquired data alone or available data alone or both may be used. See Random House Webster’s Unabridged Dictionary (2d ed. 2001) at 77 (and/or); Nobots Response Br. at 21 (so10 acknowledging for the comparing phrases); J.A. 1226. Because the “acquiring” step precedes and supplies the material for the “comparing” step, this usage for “comparing” indicates that the “acquiring” step itself can apply to data that are entirely available data (hence passive data).
Nobots relies on the specification’s repeated use of the phrase “its equivalents and verb forms” in setting forth definitions, including for “acquired data,” as quoted supra. That use is, at a minimum, not clear enough to justify Nobots’s construction in light of all that undermines it. The phrase at issue, “acquiring interest data,” is not an equivalent or verb form of “acquired data.” The middle word clearly is joined to the third word to form “interest data” as a defined unit; and disregarding it so as to affix “acquiring” directly to “data” and then treating the result as limited to “acquired data” would contradict the express definition of “interest data” as not limited to acquired data. Moreover, it is not clear just what the “its equivalent and verb forms” phrase means, and sometimes it is used where it does not seem to make sense (e.g., “available data,” “interest data,” ’008 patent, col. 3, lines 18–19, 41). A skilled artisan could only reasonably understand that the phrase was dropped in repeatedly as a stock phrase for some unspecified broadening and should not be given weight to override the otherwise-compelled meaning.
We conclude that “acquiring interest data” applies even when only passive data are obtained, and we reverse the Board’s contrary claim construction. Nobots has not disputed Google’s contention that, under the claim construction we adopt, Kitts anticipates claim 19. We therefore reverse the Board’s upholding of claim 19.
III
The decision of the Board is reversed.