When Information DoesTechnical Work: Comparing the U.S. Printed Matter Doctrine with the EPO’s Presentation-of-Information Framework
Executive Summary: This article compares the U.S. printed matter doctrine with the European Patent Office’s presentation-of-information and COMVIK frameworks, explaining that the systems overlap when semantic information is attached to otherwise known technology but are not coextensive. U.S. law asks whether a limitation is claimed for its communicative content and, if so, whether that content has the required functional relationship to the other claim elements; “technical” is therefore descriptive, not an independent U.S. requirement. EPO law more broadly evaluates mixed technical and nontechnical features and permits only features contributing to technical character to support inventive step. Through examples involving instructions, medical-device markers, data structures, human-machine interfaces, and artificial-intelligence outputs, the article shows why semantic novelty alone ordinarily cannot establish patentability. The Federal Circuit’s nonprecedential 2025 Bard decision and the EPO’s illustrative T 0642/25 decision reinforce the distinction between physical or machine-operative features and informational meaning. The resulting drafting lesson is jurisdiction-specific but practical: claims should identify how information is recognized, what structure or process acts upon it, what operation follows, and—under EPO practice—what technical purpose and effect result.
Introduction
Patent claims increasingly recite information: a warning printed on a medical device, instructions packaged with a composition, a field in a network message, a graphical prompt, a label in a machine-learning data set, or an output that causes another system to act. Whether that information can distinguish the claim from the prior art depends less on the information’s novelty than on what the information does in the claimed invention.
In the United States, that inquiry appears in the printed matter doctrine. Despite its paper-era name, the doctrine is not limited to ink on paper. It can reach information carried by visual markings, digital records, method steps, and other media when the information is claimed for its communicative content. Once a limitation qualifies as printed matter, it receives patentable weight only if it has the required functional relationship to the rest of the claim. In re DiStefano, 808 F.3d 845, 848–50 (Fed. Cir. 2015); C R Bard Inc. v. AngioDynamics, Inc., 979 F.3d 1372, 1381–82 (Fed. Cir. 2020).
European Patent Office practice reaches much of the same territory through a different route. The European Patent Convention excludes “presentations of information” only “as such,” but a claim containing technical means will ordinarily clear that threshold. The more consequential question is then inventive step: under the COMVIK approach, only features contributing to the invention’s technical character can support nonobviousness. European Patent Convention arts. 52(2)(d), 52(3), 56, Oct. 5, 1973, 1065 U.N.T.S. 199, as revised Nov. 29, 2000 [hereinafter EPC]; Case T 641/00, Two Identities/COMVIK, 2003 O.J. E.P.O. 352, headnotes I–II (Tech. Bd. App. Sept. 26, 2002).
The two systems are therefore close cousins, not twins. The frameworks overlap substantially when semantic information is attached to otherwise known technology, but they are not coextensive. The U.S. doctrine is organized around communicative content and functional relationship; COMVIK applies more broadly to claims mixing technical and nontechnical features, including business, mathematical, administrative, and other requirements that need not be presentations of information. European Patent Office, Guidelines for Examination in the European Patent Office pt. G-VII, § 5.4 (2026) [hereinafter EPO Guidelines]. In their zone of overlap, both systems resist allowing new meaning alone to manufacture patentability.
In the U.S. discussion, “technical” is used descriptively rather than as an independent legal requirement. The governing U.S. question is whether information claimed for its communicative content bears the required functional relationship to the other claim elements. EPO law separately requires a contribution to technical character for purposes of inventive step. For practitioners, the productive question is therefore not simply whether a limitation contains information, but which inquiry applies and what legally relevant relationship the claim establishes between the information and the rest of the invention.
I. The U.S. Printed Matter Doctrine
A. A doctrine of patentable weight, not claim deletion
Early printed matter cases concerned familiar articles distinguished only by words or symbols placed upon them. The Court of Customs and Patent Appeals explained that the arrangement of names in a directory could not make an otherwise old directory patentable. In re Russell, 48 F.2d 668, 669 (C.C.P.A. 1931). The doctrine later developed beyond literal printing and beyond product claims.
The modern doctrine does not permit an examiner or court simply to erase inconvenient language from a claim. Every limitation must be considered, and the claim must be read as a whole. In re Gulack, 703 F.2d 1381, 1384–85 (Fed. Cir. 1983); see also Diamond v. Diehr, 450 U.S. 175, 188–91 (1981). The doctrine instead determines what legal significance an informational limitation has in the novelty and obviousness analysis. If the limitation is printed matter and lacks the required functional relationship, its content cannot distinguish the claim from the prior art. Gulack, 703 F.2d at 1385–86.
That distinction matters. The limitation remains part of the claim for construction and infringement purposes, yet it may supply no patentable weight against an anticipation or obviousness challenge. The rule prevents an applicant from repeatedly repatenting an old product by changing the message that accompanies it. In the Federal Circuit’s memorable formulation, permitting a new instruction sheet to distinguish an otherwise old kit would allow the product to be patented indefinitely. In re Ngai, 367 F.3d 1336, 1338–39 (Fed. Cir. 2004).
B. Step one: Is the limitation claimed for communicative content?
The first question is categorical. A limitation falls within the printed matter doctrine only if it claims the content of information. DiStefano, 808 F.3d at 848. The medium is not controlling. “Printed matter” today encompasses information claimed for what it communicates, whether the information is printed, displayed, stored, spoken, or encoded. Bard, 979 F.3d at 1381–82; Praxair Distribution, Inc. v. Mallinckrodt Hospital Products IP Ltd., 890 F.3d 1024, 1032 (Fed. Cir. 2018).
The content inquiry should not be confused with a rule against all data. In DiStefano, the claim required selecting an asset for introduction into a web page but did not require that the selected asset convey any particular message. Because the limitation concerned the origin of the asset rather than the information it communicated, the Federal Circuit held that the Board had applied the doctrine too broadly. 808 F.3d at 848–50. That holding supplies a useful drafting distinction: a claim to the semantic meaning of a field is more exposed than a claim to operations performed on a field because of its defined machine role.
The same distinction explains why a visual marking, numerical sequence, or computer record is not automatically printed matter. Machine-facing data can avoid the doctrine by two analytically distinct routes. At step one, a limitation may fall outside the printed-matter category because it is claimed for its syntax, origin, organization, or machine-recognized operation rather than for communicative content. At step two, a limitation that does communicate content may nevertheless receive patentable weight when another claim element acts on it in a defined way and thereby establishes the required functional relationship. DiStefano, 808 F.3d at 848–50; Gulack, 703 F.2d at 1385–87.
C. Step two: Is there a functional relationship?
If the limitation is printed matter, the analysis turns to its relationship with the associated substrate, structure, or process. The information receives patentable weight when a new and nonobvious functional relationship exists; it does not when the physical article merely carries or displays the message. Gulack, 703 F.2d at 1385–87; DiStefano, 808 F.3d at 850.
The classic positive examples are physical. In In re Miller, volumetric indicia on a measuring cup showed fractional recipe quantities. The markings cooperated with the cup’s volumetric structure to perform a measuring function, rather than merely using the cup as a billboard. 418 F.2d 1392, 1396 (C.C.P.A. 1969). In Gulack, numbers printed around an endless band had a particular mathematical relationship because the band had no beginning or end. The substrate organized the numbers, and the numbers exploited the substrate’s structure; that interdependence required patentable weight. 703 F.2d at 1386–87.
Negative examples involve coexistence without operational interdependence. Instructions packaged with a chemical kit tell a person how to use the chemicals, but the chemicals do not depend on the instructions and the instructions do not change the kit’s structure or operation. Ngai, 367 F.3d at 1338–39. Likewise, informing a patient that taking a drug with food increases bioavailability does not transform the already-known act of taking that drug with food. King Pharms., Inc. v. Eon Labs, Inc., 616 F.3d 1267, 1278–79 (Fed. Cir. 2010).
Method claims require particular care because the relevant relationship may be causal rather than physical. In Praxair, some claims merely required providing information and evaluating whether to discontinue treatment; those informational steps did not cause a claimed action and lacked the necessary functional relationship. Other claims required discontinuing treatment “based on” the information, creating a functional relationship between the communicated result and the subsequent act. 890 F.3d at 1033–35. The difference was not that one message was more medically important. The claim language made the information operational in one instance and advisory in the other.
Bard sharpened the distinction. The claims involved vascular access ports with radiographic markers conveying that the ports were suitable for power injection. The Federal Circuit treated the markers’ identifying content as printed matter and asked whether it merely informed a person or instead interacted with the claimed device to create new functionality or cause a specific action. 979 F.3d at 1382–83. The particular apparatus claims used the marker to make the port identifiable under X-ray, which mattered to the separate eligibility analysis, but the communicative content did not automatically acquire patentable weight merely because the message concerned a medical device. Id. at 1383–85.
D. Consequences under §§ 102, 103, and 101
Printed matter most often operates in anticipation and obviousness. If nonfunctional printed matter is the only distinction over a reference, the claim may be anticipated under 35 U.S.C. § 102. If the remaining weighted limitations are an obvious combination, the content cannot rescue the claim under § 103. Ngai, 367 F.3d at 1338–39; Praxair, 890 F.3d at 1031–35. Practitioners should therefore resist describing the doctrine as solely an “obviousness rule.” Its core function is to determine whether informational content counts in either prior-art inquiry.
The doctrine can also intersect with patent eligibility, but the inquiries should not be collapsed. The Federal Circuit has held that a claim may be ineligible under § 101 when the claim as a whole is directed solely to nonfunctional printed matter and contains no additional inventive concept. Bard, 979 F.3d at 1384. In re Marco Guldenaar Holding B.V., 911 F.3d 1157, 1160–62 (Fed. Cir. 2018), illustrates the related problem where communicative dice markings and game rules supplied no technological inventive concept. Yet Bard also emphasized that printed matter and § 101 remain distinct doctrines: although some marker content was nonfunctional printed matter for prior-art purposes, the claimed radiographic identification mechanism was not merely the information itself. 979 F.3d at 1384–85. A sound analysis identifies the statutory question before borrowing language from another doctrine.
E. Digital information and machine-readable media
Digital claims expose the doctrine’s real principle. In In re Lowry, the Federal Circuit gave patentable weight to data structures stored in computer memory because the claimed data objects imposed a physical organization on memory and enabled more efficient access, insertion, and deletion. 32 F.3d 1579, 1583–84 (Fed. Cir. 1994). The objects were not claimed merely for what they told a reader; they changed how the computer operated on stored data.
By contrast, storing a song title, a batting average, or a legal conclusion on a generic medium does not ordinarily create a new functional relationship with the medium. The current USPTO formulation reflects this divide: programming that performs a function with respect to the computer may be functionally related, while information stored only for its communicative content may not be. Manual of Patent Examining Procedure § 2111.05 (9th ed. Rev. 01.2024).
The practical lesson is to claim the processing consequence, not simply the informational label. A field “indicating authorization” is vulnerable if a human merely reads it. A field that a controller validates, uses to select a protocol, and relies on to enable a defined machine operation is far better positioned. The latter claim identifies a closed technical relationship among the data, the processor, and the resulting state change.
II. The EPO’s Presentation-of-Information Framework
A. Article 52: exclusion only “as such”
Article 52(2)(d) of the EPC lists presentations of information among subject matter not regarded as inventions, while Article 52(3) limits the exclusion to the listed subject matter “as such.” EPC arts. 52(2)(d), 52(3). A pure claim to the content of a message may therefore face an exclusion. But a claim to a display, computer, medical device, or other technical means ordinarily has technical character as a whole and passes the initial Article 52 inquiry, even if much of the claim concerns information.
That threshold is important but rarely the end of the analysis. The EPO Guidelines expressly use a kit containing a composition and instructions as an example: the composition supplies technical character, so the claim is not excluded as a presentation of information as such. The instructions must still be evaluated for novelty and inventive step, and they cannot support inventive step unless they contribute to a technical effect serving a technical purpose. EPO Guidelines, supra, pt. G-II, § 3.7.
For U.S. readers, the institutional distinction matters: the EPO Guidelines guide examination practice but do not bind the Boards of Appeal. A Board may consider the Guidelines while applying the EPC and Board case law. Case T 0642/25, Training set augmentation/FEATURESPACE, ¶ 6.4.2 (EPO Tech. Bd. App. Apr. 22, 2026).
This sequencing differs from common U.S. phrasing. U.S. law often speaks of giving a limitation “no patentable weight.” EPO practice first recognizes that a mixed claim may be an invention as a whole, then separates the features that contribute to technical character when applying Article 56.
B. COMVIK: only a technical contribution can support inventive step
The governing framework is the COMVIK approach. A claim may legitimately contain both technical and nontechnical features, even if the nontechnical features dominate. But only features that contribute to the claim’s technical character can support an inventive step. Two Identities/COMVIK, 2003 O.J. E.P.O. 352, headnote I, ¶¶ 4–6. A nontechnical requirement may be included in the objective technical problem as a constraint given to the skilled person; ingenuity in choosing that requirement is not itself credited as the technical solution. Id. headnote II, ¶ 7.
In practice, the examiner identifies the differences from the closest prior art, determines the technical effects of those differences in the context of the claim as a whole, and separates features that make a technical contribution from those that do not. If there is no difference at all, Article 54 novelty fails. If there is a difference but it makes no technical contribution, Article 56 inventive step fails because there is no technical contribution to the prior art. EPO Guidelines, supra, pt. G-VII, § 5.4.
This is one of the most useful points of comparison with U.S. practice. A newly worded message may be a literal difference for EPO novelty, yet still be incapable of supporting inventive step. In the United States, nonfunctional printed matter may receive no patentable weight under either § 102 or § 103. The destination is often similar, but the doctrinal route—and sometimes the formal ground of rejection—is different.
C. Cognitive content, “what” is presented, and “how” it is presented
The EPO analyzes both the cognitive content of information—what is presented—and its form or arrangement—how it is presented. EPO Guidelines, supra, pt. G-II, § 3.7. Neither category is automatically technical. A new business message, aesthetic layout, or arrangement that merely makes information more appealing or easier to understand normally cannot support inventive step. The alleged benefit cannot rest only on subjective interests, preferences, or cognitive abilities.
There are, however, two significant pathways to technical contribution. First, the content can reflect a dynamically changing internal state of a technical system and enable the user to operate that system properly. A warning generated from an automatically detected machine condition may contribute technically when it prompts corrective interaction; static instructions describing how a device could be operated generally do not. EPO Guidelines, supra, pt. G-II, § 3.7.
Second, the manner of presentation can credibly assist a user in performing a technical task through a continued or guided process of human-machine interaction. The assistance must be objectively, reliably, and causally linked to the claimed presentation feature. Id. In Case T 643/00, the Board recognized that simultaneously presenting several low-resolution images and permitting selection of a high-resolution version could help a user perform the technical task of searching and retrieving stored images more efficiently. Case T 643/00, Searching Image Data/Canon, catchword (EPO Tech. Bd. App. Oct. 16, 2003).
That standard is more exacting than saying that a display is “user friendly.” The claim and specification should identify the technical task, the information that guides the task, the interaction it causes, and the objective technical effect. A surgeon’s display that continuously presents detected implant orientation in a manner enabling a more precise correction is a stronger case than a static diagram explaining the procedure. A controller display that shows a detected overload and guides a reset sequence is stronger than a manual listing possible error codes.
D. Functional data versus cognitive data
The EPO’s treatment of data structures closely parallels the U.S. distinction between Lowry-type machine functionality and communicative content. Functional data can contribute to technical character when its structure has an intended technical use and causes a technical effect in that use. Cognitive data, by contrast, is relevant because of its meaning to a human user. EPO Guidelines, supra, pt. G-II, § 3.6.3.
In Case T 1194/97, a record carrier stored coded picture data whose line synchronizations, line numbers, and addresses were defined by the operation of the picture-retrieval system. The Board held that such functional data was not a presentation of information as such because it inherently reflected technical features of the system in which the carrier operated. Case T 1194/97, Data Structure Product/Philips, 2000 O.J. E.P.O. 525, headnotes I–II (Tech. Bd. App. Mar. 15, 2000). Similarly, an index file can be technical when it controls how a computer searches for records. Case T 1351/04, File Search Method/Fujitsu, catchword (EPO Tech. Bd. App. Apr. 18, 2007).
The same electronic object can contain both kinds of data. A message header automatically recognized by a receiving system may determine how content is assembled and processed, while the message body remains cognitive content for the recipient. EPO Guidelines, supra, pt. G-II, § 3.6.3. That distinction should influence claim architecture: describe the field’s machine-recognized syntax, validation, processing path, and effect rather than relying only on its semantic name.
III. Overlapping Boundary, Different Legal Machinery
The frameworks’ overlapping treatment of information can be mapped as follows:
Issue | United States | European Patent Office |
Scope | Limitations claimed for the content of information | Mixed technical and nontechnical features generally; presentations of information are one category |
Initial classification | Is the limitation claimed for the content of information? | Is the claim directed to a presentation of information “as such,” or does it contain technical means? |
Operative test | Is the printed matter functionally related to the substrate, structure, or process? | Does the feature contribute to technical character by producing a technical effect serving a technical purpose? |
Human-facing information | Usually weak if it merely informs; stronger if the claim makes it cause a specific action or cooperate with the article | Usually weak if benefit is cognitive or subjective; stronger if objectively and causally linked to a guided technical task |
Machine-facing data | Stronger when organization or content changes computer operation | Stronger when functional data controls or reflects operation of a technical system |
Prior-art consequence | Nonfunctional printed matter may receive no patentable weight under §§ 102 and 103 | A nontechnical difference may establish literal novelty, but cannot support Article 56 inventive step |
The U.S. “substrate” concept and the EPO “technical system” concept should not be treated as synonyms. The U.S. doctrine retains the language of a relationship between printed matter and an associated physical or process context. EPO analysis is more explicitly effect-centered: it asks whether the feature contributes to solving a technical problem in the claim’s context. But the same factual evidence often matters in both systems—interdependence, automatic processing, causal action, measurable system behavior, and a result that does not depend merely on a person appreciating meaning.
The treatment of human action illustrates the nuance. U.S. cases can recognize a functional relationship where the claim expressly requires an action “based on” information, as in the treatment-discontinuation claims in Praxair. 890 F.3d at 1034–35. EPO practice asks a differently framed—and often more demanding—question in human-facing cases: the presentation must objectively, reliably, and causally assist performance of a technical task through continued or guided interaction. EPO Guidelines, supra, pt. G-II, § 3.7. A claim requiring a user to choose an investment after reading a risk score may therefore remain nontechnical, while a claim guiding an operator through a sensor-calibration sequence may receive a different analysis. That comparison is limited to human-facing cases; EPO technical contribution may also arise without a claimed human action when information reflects an automatically detected, dynamically changing internal state of a technical system. Id.
IV. Four Recurring Claim Patterns
A. The product-plus-instructions kit
Consider a claim to a known reagent kit packaged with instructions directing a newly discovered use. In the United States, Ngai makes the risk plain: if the instructions and reagents do not depend on one another, the content cannot distinguish the old kit. 367 F.3d at 1338–39. At the EPO, the reagent supplies technical character and avoids exclusion under Article 52, but the new instructions ordinarily make no technical contribution and cannot establish inventive step. EPO Guidelines, supra, pt. G-II, § 3.7.
The stronger claim is usually directed to the new use itself, if the governing law permits that claim form and the use is otherwise patentable, or to a kit whose components are structurally adapted to carry out the use. Merely placing a new instruction sheet next to an old product is the weakest common fact pattern in both jurisdictions.
B. The marker that identifies a medical device
Suppose a port includes a marker indicating that it tolerates a specified injection pressure. A text label read by a clinician is classic communicative content. A marker whose geometry or material produces a defined radiographic signature, detected by imaging equipment to identify the port, presents a stronger technical story. Under U.S. law, Bard requires separate attention to whether the marker’s message is functionally related to the port and whether the claimed identification mechanism supplies more than the abstract information itself. 979 F.3d at 1382–85. On the claims before it, the court held that self-identification alone did not create new functionality and that the method claims lacked language causally tying identification to power injection. Id. at 1383–84.
A nonprecedential 2025 follow-on decision after remand further separated the marker’s physical detectability from its semantic meaning. The Federal Circuit explained that a radiographic marker received patentable weight only to the extent it was a radiographically discernible structural feature—not because a clinician attributed identifying meaning to the marker or to radiographic letters. It likewise held that the letters’ communicative significance could not distinguish the prior art. C.R. Bard, Inc. v. AngioDynamics, Inc., No. 2023-2056, slip op. at 5–6, 9 (Fed. Cir. Dec. 15, 2025) (nonprecedential). The disposition is persuasive clarification, not controlling new law.
Under EPO principles, a machine-detectable radiographic signature is not merely a human-facing presentation. Its generation, detection, and use may be technical features if the claim connects them to imaging or device control. The drafter should claim that chain: a marker structure that produces a specified signal; a detector that extracts it; a processor that verifies a device property; and, where supported, a controller that enables or limits operation in response.
C. The data structure or message field
A record that merely stores the statement “authorized” is vulnerable. A record with a field whose location and format are automatically validated, whose value selects a protocol, and whose verification changes the state of a controlled device is much stronger. In the United States, that distinction aligns with Lowry and DiStefano: machine organization and operation are different from content claimed solely for its meaning. Lowry, 32 F.3d at 1583–84; DiStefano, 808 F.3d at 848–50. At the EPO, it aligns with functional data under Philips and the Guidelines. Data Structure Product/Philips, 2000 O.J. E.P.O. 525, headnotes I–II; EPO Guidelines, supra, pt. G-II, § 3.6.3.
The label assigned to the field will not do the work. The claim should recite the machine relationship: where the field appears, how it is recognized, what processing it controls, and which technical behavior follows.
D. The recommendation followed by action
Claims often recite generating a recommendation and displaying it to a user. That formulation leaves the output on the communicative side of the line. A stronger claim requires a defined actor to use the result in a claimed operation. Praxair shows why causal language matters in the United States. 890 F.3d at 1033–35. At the EPO, however, adding “based on” is not talismanic. The action must be part of a technical task, and the informational feature must be objectively and causally linked to the technical effect.
Thus, “displaying a recommendation to reduce energy use” remains weak. “In response to a detected phase imbalance, displaying a sequenced switching instruction and receiving confirmations that advance a controller through a safe reconfiguration state machine” identifies a guided technical interaction. Strong claims describe the loop, not merely the advice.
V. Drafting and Prosecution Strategies
1. Begin with the technical consequence
Before drafting the informational limitation, complete this sentence: “Because the system receives or presents this information, the claimed system ______.” If the answer is only “the user knows,” the limitation is exposed. Better answers describe a changed memory operation, protocol selection, device state, measurement, error recovery, resource use, or guided technical interaction.
2. Claim the causal chain
Use claim language that connects information to action: detect, encode, validate, select, control, inhibit, enable, reconfigure, or advance in response to. Identify the actor and the result. “A status field indicating a fault” is less useful than “a controller validating the status field and, in response to a fault value, isolating the affected channel before initiating a recovery sequence.” The second formulation makes the claimed dependency visible.
3. Define the relevant technical context
For U.S. purposes, identify the article, structure, memory organization, or process with which the information cooperates. For EPO purposes, identify the technical task and the technical effect. The specification should explain why the relationship is not arbitrary and why the result follows from the claimed features rather than from a user’s subjective preference.
4. Build objective evidence into the disclosure
EPO technical-effect arguments become more credible when the application supports measurable results: reduced retrieval operations, lower latency, fewer erroneous inputs, improved control stability, reduced memory overhead, or a more precise surgical correction. The same evidence can strengthen the U.S. account of a genuine functional relationship. Draft the application with comparative embodiments and objective metrics before those facts are needed in prosecution.
5. Separate human-readable and machine-operative embodiments
Many inventions use the same underlying information in two ways. One field may be displayed to a person and also parsed by a controller. The specification and claims should not blur those roles. Describe separate embodiments and claim paths: a human-facing presentation with guided interaction; a machine-facing encoding with defined processing; and, where appropriate, a closed-loop method that links both.
6. Prepare for different rejection labels
A U.S. examiner may cite the printed matter doctrine in a § 102 or § 103 rejection and assert that the informational distinction deserves no weight. The response should address both steps: explain why the limitation is not claimed for communicative content, or, if it is, identify the new functional relationship. An EPO examiner may acknowledge literal novelty but omit the informational difference from the technical problem or treat it as a nontechnical constraint under COMVIK. The response should identify the claimed technical effect, show that it is caused across the claim’s scope, and formulate the objective technical problem accordingly.
VI. AI Systems Put the Distinction Under Pressure
Artificial-intelligence claims frequently recite labels, prompts, explanations, rankings, recommendations, confidence values, and generated plans. Those features can look technologically sophisticated while remaining informational in the relevant legal sense. A model output does not become functionally or technically significant merely because a neural network generated it.
An illustrative recent Board of Appeal decision applies these principles to machine-learning training data. In Case T 0642/25, the Board considered a method that generated synthetic transaction samples and labels for training an anomaly-detection model. It held that neither model training nor improved classification accuracy was technical per se where the classifier served the nontechnical purpose of identifying anomalous financial transactions. The labels represented cognitive information about the training data, not functional data within Philips. The official record designates the decision “No distribution,” provides no headnote or catchword, and shows that it was placed online on July 29, 2026; it is best read as a recent application of existing principles rather than an independent EPO-wide doctrinal development. Case T 0642/25, Training set augmentation/FEATURESPACE, ¶¶ 6.2–6.4.2 (EPO Tech. Bd. App. Apr. 22, 2026).
That decision does not mean that training data or model outputs can never contribute technically. It means the claim must connect them to a technical purpose and effect. Training an image-processing model through transformations tied to the digital representation of images presents a different case from assigning business labels to transaction records. Likewise, a model output that directly sets an actuator trajectory, controls a radio parameter, or selects a memory-management operation is different from a score displayed for a person’s consideration.
The same discipline should guide U.S. drafting. A generated explanation intended for a human is likely communicative content. A token sequence defined by machine-recognized syntax and claimed for its role in controlling a downstream processor may fall outside the printed-matter category at step one. Alternatively, where the sequence also communicates content, the claimed parsing, validation, and state transition may establish the required functional relationship at step two. For agentic systems, “generating a plan” is weak standing alone; validating plan steps against machine constraints and executing approved commands to alter a controlled system supplies the operational relationship.
As AI systems distribute functions among models, people, and tools, the identity of the consumer becomes increasingly important. Is the information consumed cognitively by a person, syntactically by software, or physically by a device? Does the consumer merely receive the information, or does the claim require a particular operation because of it? Those questions are becoming the modern form of the substrate and technical-effect inquiries.
Conclusion
The printed matter doctrine and the EPO’s presentation-of-information framework share an overlapping concern: patentability should not turn on a newly composed message attached to otherwise old technology. But their legal tests are distinct. U.S. law asks whether information claimed for its communicative content has the required functional relationship to the other claim elements. EPO law applies the broader COMVIK framework and asks whether a feature contributes to technical character for inventive-step purposes.
The drafting mandate is therefore concrete but jurisdiction-specific. In a U.S. claim, identify what receives the information, how the information cooperates with structure or process, and what defined operation or action depends on it. In an EPO claim, identify the technical purpose and the technical effect caused across the claim’s scope. In neither system should semantic novelty be expected to carry the claim by itself.
